Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
University of South Carolina University of South Carolina
Scholar Commons Scholar Commons
Theses and Dissertations
Summer 2019
Two Essays on Market Entry and Exit: Empirical Evidence from Two Essays on Market Entry and Exit: Empirical Evidence from
Airline Industry Airline Industry
Sina Aghaie
Follow this and additional works at: https://scholarcommons.sc.edu/etd
Part of the Business Administration, Management, and Operations Commons
Recommended Citation Recommended Citation Aghaie, S.(2019). Two Essays on Market Entry and Exit: Empirical Evidence from Airline Industry. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/5421
This Open Access Dissertation is brought to you by Scholar Commons. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of Scholar Commons. For more information, please contact [email protected].
TWO ESSAYS ON MARKET ENTRY AND EXIT: EMPIRICAL EVIDENCE FROM
AIRLINE INDUSTRY
by
Sina Aghaie
Bachelor of Science
Sharif University of Technology, 2008
Master of Business Administration
Sharif University of Technology, 2011
Submitted in Partial Fulfillment of the Requirements
For the Degree of Doctor of Philosophy in
Business Administration
Darla Moore School of Business
University of South Carolina
2019
Accepted by:
Rafael Becerril Arreola, Major Professor
Carlos J.S. Lourenço, Major Professor
Satish Jayachandran, Committee Member
Chen Zhou, Committee Member
Charles H. Noble, Committee Member
Cheryl L. Addy, Vice Provost and Dean of the Graduate School
ii
© Copyright by Sina Aghaie, 2019
All Rights Reserved.
iii
DEDICATION
To my mother, who has always loved me unconditionally and whose good examples have
taught me to work hard for the things that I aspire to achieve. To my brother, for his
unwavering support throughout my academic career. To my best friends, Stella,
Anooshiravan, Ali Kazi and Behrang whose unconditional support and happy nature
remind me to smile when things get tough. To Dr. Afshin, who has always been a constant
source of support and encouragement. And finally, to Prof. Ming-jer Chen, whose seminal
works got me interested in competitive dynamics, inspired my research and helped me in
all the time of working on my dissertation. I could not have imagined having better
guidance for my Ph.D. study.
iv
ACKNOWLEDGEMENT
I am indebted to my mentors Dr. Carlos Lourenço and Dr. Charles Noble for all of their
trust, all of their support, and their insightful feedback on my work. I extend my heartfelt
gratitude to Carlos and Charles who devoted their time to work with me, teach me and help
me develop different pieces of my Ph.D. research work. Carlos has been patient with me
during the challenging years of my PhD and taught me through my very first steps into
research. I am simply thankful to him in so many ways, for so many things.
I am grateful to my committee members, Dr. Satish Jayachandran, Dr. Rafael
Becerril Arreola, and Dr. Chen Zhou. I would like to thank all of the professors I had chance
to attend their courses and learn from them, in particular, professors Dr. Priyali Rajagopal,
Dr. Kartik Kalaignanam, Dr. Orgul Ozturk, and Dr. Mehdi Sheikhzadeh.
I could not do this without the support of my friends. Many thanks to the beautiful
mind, Ghahhar for providing good feedback on my projects. Thank you, Taehoon, Gustavo,
Ruoou, Arka, Amir, Sotos, Aboulghasem, Hesam, Mehdi, Mohammad, Nima, Danial,
Javad, Soheil, and Omid for all of your support. Your friendship makes my life a wonderful
experience.
Many thanks must go also to my second parents Nayyereh and Ali Mohammad, for
their advice, undying support, and their unwavering belief that I can achieve so much.
Finally...I want to thank all of those people who, once upon a time, were my teachers and
who helped me begin this long journey. Their names are too numerous to mention, but
many of them inspired me to continue learning and sharing with others.
v
ABSTRACT
The proliferation of low-cost competitors has increasingly eroded incumbent firms’ market
shares and profitability in recent decades. However, incumbents are still uncertain about
how to handle this new challenge. The two essays in this dissertation aim to contribute to
the marketing strategy and competitive dynamics literature by exploring the link between
incumbents’ marketing-mix activities and low-cost rival’s market entry, exit, and the threat
of entry decisions.
In the first essay, I study a common and important phenomenon – the marketing
tactics that incumbent firms employ to drive new low-cost entrants out of the market.
Specifically, I investigate how incumbents’ price, service quality, and service convenience
influence an entrant’s market exit, and how this influence may change over time. The
hypotheses are tested on a rich, longitudinal dataset from the US airline industry between
1997 and 2016. I estimate challengers’ time-to-exit using a split population hazard model
that accounts for challengers that ‘never’ exit. Instead of homogeneous results, I find that
the magnitude and direction of the effects vary over time. For instance, a substantial price-
cut initially delays but will later accelerate an entrant’s exit timing. I suggest that managers
should take into account the type (price vs. quality), timing (sooner vs. later after entry),
and intensity (more vs. less) of defensive responses to a new low-cost entrant.
vi
When a firm makes an action that takes it closer to a market, incumbent firms would
profit from knowing whether such threat is to be taken seriously – and one that incumbents
could do something about – or a bluff – that incumbents could ignore. Thus, in the second
essay, I estimate the probability of a serious (vs. a bluff) threat as a function of market
characteristics as well as the characteristics of the potential entrant and those of incumbents
and the structure of their market network. In line with the awareness-motivation-capability
framework, I argue that a threat is more (less) likely to be serious (bluff) when the potential
entrant has the motivation to enter the market as well as the capability of doing so. This
study provides insights for managers of incumbent firms on how to more effectively and
efficiently allocate limited marketing resources over time to defend ‘their’ markets – or do
nothing – in the face of a rival’s threat of entry.
vii
TABLE OF CONTENTS
DEDICATION ................................................................................................................... iii
ACKNOWLEDGEMENT ................................................................................................. iv
ABSTRACT .........................................................................................................................v
LIST OF FIGURES ........................................................................................................... ix
LIST OF TABLES ...............................................................................................................x
ESSAY 1: REPELLING INVADERS: USING MARKETING TACTICS TO THWART
LOW-COST ENTRANTS ...................................................................................................1
THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT ...............4
EMPIRICAL ANALYSIS .............................................................................................13
RESULTS ......................................................................................................................21
GENERAL DISCUSSION ............................................................................................28
ESSAY 2: BLUFF OR REAL? HELPING INCUMBENTS RECOGNIZE HOW A
REALLY THREATENING FIRM LOOKS LIKE ...........................................................46
THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT .............49
DATA, INDUSTRY CONTEXT, AND THREAT CLASSIFICATION ......................53
EMPIRICAL ANALYSIS .............................................................................................57
RESULTS ......................................................................................................................67
GENERAL DISCUSSION ............................................................................................72
REFERENCES ..................................................................................................................83
APPENDIX A: DESCRIPTIVE STATISTICS AND ROBUSTNESS CHECKS ............96
APPENDIX B: FULL LIKELIHOOD FUNCTION IN THE SPLIT-POPULATION
HAZARD MODEL ..........................................................................................................102
viii
APPENDIX C: 2SRI METHOD ......................................................................................106
APPENDIX D: VARIABLE MEASUREMENTS ..........................................................109
APPENDIX E: TYPES OF MISSINGNESS ...................................................................110
ix
LIST OF FIGURES
FIGURE 1.1: CONCEPTUAL FRAMEWORK .............................................................. 37
FIGURE 1.2: DISTRIBUTION OF MARKET EXITS OVER TIME ............................. 38
FIGURE 1.3: TIME-DEPENDENT EFFECTS OF INCUMBENTS MARKETING
TACTICS ON A CHALLENGER’S EXIT TIMING ...................................................... 39
FIGURE 2.1: CONCEPTUAL FRAMEWORK .............................................................. 77
FIGURE 2.2: THREAT ESTABLISHMENT .................................................................. 78
FIGURE A.1: INCUMBENTS’ MARKETING-MIX TACTICS BEFORE AND AFTER
A CHALLENGER’S ENTRY .......................................................................................... 96
x
LIST OF TABLES
TABLE 1.1: LITERATURE REVIEW OF ANTECEDENTS OF MARKET EXIT AND
STUDY CONTRIBUTION .............................................................................................. 40
TABLE 1.2: VARIABLE OPERATIONALIZATION .................................................... 43
TABLE 1.3: SPLIT-POPULATION MODEL RESULTS ............................................... 44
TABLE 1.4: ROBUSTNESS CHECKS ........................................................................... 45
TABLE 2.1A: SERIOUS THREAT VS. BLUFF THREAT CLASSIFICATION .......... 79
TABLE 2.1B: DATA STRUCTURE ............................................................................... 79
TABLE 2.2: DESCRIPTIVE STATISTICS AND CORRELATION MATRIX ............. 80
TABLE 2.3: RESULTS ACROSS DIFFERENT IMPUTATION TECHNIQUES ......... 81
TABLE 2.4: RESULT WITH MULTIPLE IMPUTATION CHAINED EQUATIONS . 82
TABLE A.1:DESCRIPTIVE STATISTICS AND CORRELATION MATRIX ............. 97
TABLE A.2: ROBUSTNESS CHECK – ENTRY DEFINITION ................................... 98
TABLE A.3: ROBUSTNESS CHECK-REACTION DEFINITION ............................... 98
TABLE A.4: ROBUSTNESS CHECK-DISTRIBUTION ASSUMPTION ................... 100
TABLE A1.5: ROBUSTNESS CHECK-DISTRIBUTION
ASSUMPTION (CONT’D) ............................................................................................ 101
1
ESSAY 1: REPELLING INVADERS:
USING MARKETING TACTICS TO THWART LOW-COST
ENTRANTS1
“We’ve finally reached the point, perhaps, where [low cost carrier] penetration may be fatal.” – David
Grizzle, Senior Vice President, Continental Airlines.2
Incumbent firms across many industries face the challenge of an increasing threat: the entry
of rivals into ‘their’ markets. These market entries disrupt incumbents, damages their
margins, and may dramatically change the rules of the game (Luoma, et al. 2018; Spann,
Fischer, and Tellis 2014). While incumbent firms can take some comfort in academia and
industry reports indicating a high chance of exit after an entry (Horn, Lovallo, and Viguerie
2005; Luoma, et al. 2018; Robinson and Min 2002), they cannot simply ‘wait and see’, as
allowing a new entrant to survive and eventually thrive can have devastating effects.
Instead, they take an active role towards firms entering ‘their’ markets, especially when
new entrants are ‘low-costs.’3
Past research on the antecedents of the market entry failure has devoted a lot of
attention to the market- and new-entrant’s characteristics such as overall expected demand,
industry concentration (Dunne et al. 2013; Van Kranenburg, Palm, and Pfann 2002), firm
age and size, entry timing, pre-entry experience and knowledge, multi-market contact, and
1 Aghaie, Sina., Carlos Lourenço, and Charles Noble. To be submitted to Marketing Science 2 Source: “Low cost airlines put the crunch on biggest carriers,” The Wall Street Journal, June 19, 2002. 3 One of many similar examples involves EasyJet, one of Europe’s biggest and most successful low-cost
airlines. In early 2017 EasyJet announced it would stop flying the Lisbon-Ponta Delgada route in Portugal,
two years after having moved in. According to its managers, despite the growing demand in that market, the
low-cost airline left because it could not guarantee its service standards, namely in terms of flight frequency,
though customers argue the truth is it could no longer cope with low prices practiced by incumbents. In other
words, the marketing tactics of incumbents at some point can drive the low-cost entrant out of the market.
Source: http://theportugalnews.com/news/easyjet-leaves-the-azores-ryanair-launches-promotion/41541
2
both mode and order of entry, and strategic fit (Boeker et al. 1997; Gatignon, Robertson,
and Fein 1997; Homburg et al. 2013; Johnson and Tellis 2008; Papyrina 2007; Sinha and
Noble 2008; Sousa and Tan 2015). Surprisingly, most studies either ignored the link
between incumbents’ marketing tactics and new entrants’ time-to-exit or implicitly
assumed that rival’s characteristics and activities (i.e., incumbent firms in the market) have
no impact on the new entrant’s survival. Moreover, except for a few studies (Geroski, Mata,
and Portugal 2010 and Nikolaeva 2007), prior literature usually explored the time-invariant
effects of market exit drivers and remained mostly silent about how those effects may
change over time.
In this paper, we take a step at addressing these gaps by investigating how
incumbents’ price- and non-price marketing arsenals influence the new entrant’s market
exit, and how these influences may evolve over time. More specifically, we link the time-
to-exit of a new low-cost market entrant to incumbents’ price, service convenience, and
service quality. Drawing on the related notions of action irreversibility (Chen and
MacMillan 1992; Chen et al. 2002) and information economics (Connelly et al. 2011;
Panagopoulos et al. 2018; Prabhu and Stewart 2001; Talay, Akdeniz, and Kirca 2017), we
predict that the ability of incumbents’ price-cuts to repel a low-cost newcomer4 actually
grows over time. On the other hand, we expect that an incumbent’s better service
convenience and higher service quality accelerates a newcomer’s exit time regardless of
the newcomer’s time in the market.
We focus on the entries by the low-cost firms in markets previously dominated by
premium (vs. low-cost) incumbent firms. Low-cost entrants proliferate at a higher rate
4 Note that we use “newcomer”, “new entrant”, and “challenger” interchangeably.
3
today than they did a decade ago (Ryans 2009), and in a range of industries, from grocery
retailing (e.g., Wal-Mart and, more recently, Aldi supermarkets in the US) to the airline
industry (e.g., Southwest in the US or EasyJet in Europe), to consumer technology (e.g.,
Huawei’s rapid global expansion in the less expensive smart phone market). We test our
hypotheses empirically on an extensive, multi-market longitudinal dataset from the US
airline industry. The volatile demand and a very competitive nature of the airline industry
make it an attractive context for this research. In this industry, airlines compete with each
other through price-cutting, service convenience (e.g., flight frequency) and service quality
differentiation (Ethiraj and Zhou 2019). Additionally, budget airlines expansion has been
frequently cited as one of the primary causes of premium airlines’ financial crisis (Ito and
Lee 2003) 5, so, no wonder those in the airline industry, might see “a low-fare carrier
coming into their turf like getting cancer” and, sooner or later, they “want to cut it out.”6
Our results suggest that price-cuts although are the easiest and fastest way of
responding to a new low-cost entrant, but at the same time, may not be the most efficient
tools to quickly drive a new entrant out of the market and service-based strategies may be
better suited for that task. For managers of incumbent firms, our findings may help to
implement effective marketing tactics over time to repel new (low-cost) entrants. We
contribute to prior literature on the antecedents of firm survival (Homburg et al. 2013;
Johnson and Tellis 2008; Lieberman, Lee, and Folta 2017; Robinson and Min 2002; Wang,
Chen, and Xie 2010) by introducing a broader set of marketing factors that impact a new
entrant’s survival. Namely, we investigate price, service convenience, and service quality
5 See also Informational Brief of United Airlines, Inc., In the United States Bankruptcy Court For the Northern
District of Illinois, December 9, 2002. 6 See https://www.wsj.com/articles/SB1031516620409380155
4
side-by-side, instead of focusing on just price (e.g., Dixit and Chintagunta 2007). Also, in
contrast to previous research in marketing that has studied low-cost entrants’ time-to-exit
in a static environment, we look at potential changes of marketing effects over time, which
may shed light on mixed findings in the literature regarding the effect of incumbents’ prices
on market exit (Dixit and Chintagunta 2007; Gatignon, Robertson, and Fein 1997).
The paper is organized as follows. First, we establish the theoretical background on
market exit drivers in the context of a new low-cost entrant. Second, we develop the
conceptual framework and predictions relating incumbents’ marketing tactics, namely
those related to price, service convenience, and service quality to the newcomer’s time-to-
exit. Third, we discuss the empirical modeling and estimation strategies and describe the
airline industry data and the operationalization of the different variables used. Finally, we
present the results of the study and consider implications for advancing marketing practice
and future research.
THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT
Incumbent firms have relatively few general weapons with which to fend off invaders,
particularly those that may operate more efficiently or be more deep-pocketed to weather
a pricing war. Customer loyalty, traditionally the most powerful sustainable advantage of
incumbents, is declining as customers become more transactional and even seek new
brands (Dawes et al. 2015; Lamey 2014; Umashankar et al. 2017; Wieseke et al. 2014)7.
Beyond the strategically uninspired approach of price warfare, we believe certain service
tactics may be key to repelling new, low-cost entrants, and explore such approaches here
7 https://www.brandchannel.com/2011/09/12/brand-loyalty-and-the-recession-its-all-about-passionistas-vs-
frugalistas/
5
(Lusch, Vargo, and O’brien 2007; Obeng et al. 2016). We first consider the history of
research exploring the antecedent of market exit in Table (1.1). This overview highlights
the novel contributions of this study in considering various price and service-based tactics
with time-variant effects considered.
As we consider the potential strategic benefits of pricing and service tactics in the
face of new entrants, we view incumbent actions as market information signals that can
create (or alleviate) uncertainty for the new entrant and influence its competitive behaviors.
Thus, the information economics perspective guides this research, particularly the
underlying principles of: (1) information asymmetry, and (2) signaling effects.
Information asymmetry. In the business world, exchange parties often have
information differentials, otherwise known as an asymmetry (Panagopoulos et al. 2018).
In our context, information asymmetry occurs when an incumbent has more and/or better
information relative to a challenger about a market. We expect that a challenger’s
information disadvantage triggers uncertainty regarding the future outlook of the market.
As a result, the challenger firm constantly seeks information to reduce its uncertainty. Prior
studies have indicated that competitors are one of the main sources of information and their
activities contain embedded signals. However, the newcomer must process and evaluate
these signals in order to resolve its uncertainty (Hsieh, Tsai, and Chen 2015; Luoma, et al.
2018; Prabhu and Stewart 2001).
Signaling. A market signal can be “any action by a competitor that provides a direct
or indirect indication of its intentions, motives, goals, or internal situation” (Porter 1980,
p.75). Traditionally, signaling theory has focused on two parties – the sender and receiver
of the signal. Senders are seen as the entity of interest and possess unique information (i.e.,
6
are the more informed party), while receivers stand to benefit from the sender’s signal and
seek the information behind it (Vasudeva, Nachum, and Say 2018). The focus of research
applying this theory has been on how sender attributes and intentions are inferred by the
receiver in the absence of unequivocal information (Connelly et al. 2011). Fundamentally,
signaling theory attempts to explain how information-disadvantaged firms use “credible”
signals to reduce information asymmetry and competitive disadvantage (Steigenberger and
Wilhelm 2018; Talay, Akdeniz, and Kirca 2017). A credible signal requires an apparent
commitment to that course of action and action “irreversibility” level reflects the sender’s
commitment to that competitive action. Thus, the notion of action irreversibility is central
in establishing credibility (Wang and Xie 2011).
Action irreversibility. Action is irreversible to the degree that, once undertaken, it
is hard to change it in the future (Steen 2016). Perceived irreversibility can signal a
commitment to an impending action and the unlikelihood it will be revoked. Prior studies
have found that the degree of perceived irreversibility of competitive actions will shape
rival behaviors (Chen and MacMillan 1992; Chen et al. 2002). According to Michael
Porter: “Perhaps the single most important concept in planning and executing offensive
and defensive competitive actions is the concept of commitment ... The persuasiveness of
a commitment is related to the degree to which it appears binding and irreversible” (1980:
100-101). The irreversibility of the incumbent’s action can significantly impact the
challenger’s response behavior because it acts as a strong signal of the tenacity of a
defender (Chen and MacMillan 1992).
Irreversibility continuum. Prior studies have considered the irreversibility level as
a spectrum, ranging from highly reversible to the highly irreversible. Highly reversible
7
actions can be reversed costlessly at any time (Chen et al. 2002). On the other end of this
continuum are those actions that are highly irreversible because once the firm launches the
action, it cannot get back to where it was before (Ghemawat 2016). Usually, marketing
activities fall within this continuum where some actions are more readily reversible (e.g.,
price changes), some have a moderate level of irreversibility (e.g., promotions) and others,
typically involving large investments, are more irreversible (e.g., mergers and acquisitions,
development of new products). Given the binding nature of the irreversible actions, the
vast majority of competitive responses in practice are more reversible in nature (Chen et
al. 2002) so that the incumbent can maintain some strategic flexibility (Trigeorgis and
Reuer 2017; Steenkamp et al. 2005). Built directly from the aforementioned tenets, the
conceptual framework is predicated on an evolutionary view of a firm’s time-to-exit as a
function of not only its own actions and characteristics but also, and decisively, those of
incumbent competitors (Homburg et al. 2013; Reibstein and Wittink 2005), namely related
to price, service convenience and service quality. We draw on these insights to predict how
incumbents’ marketing activities affect the new entrant’s time-to-exit.
Incumbents’ Marketing Tactics in the Face of New Entrants
Incumbents usually adjust their marketing-mixes when faced with competitors entering
their markets (Hauser and Shugan 2008; Kuester, Homburg, and Robertson 1999; Prince
and Simon 2014; Shankar 1997; Simon 2005). Along with pricing tactics (Goolsbee and
Syverson 2008; Luoma, et al. 2018), incumbent firms may use their services as strategic
weapons to protect themselves from a new entrant. For instance, Obeng et al. (2016) found
that incumbents with better services are more likely to withstand new competitive threats
than those with fewer services. Thus, in this research, we focus on incumbents’ pricing and
8
service strategies as two main marketing tactics that incumbents will use in response to the
new entrants.
Incumbents’ price and the challenger’s exit timing. Perhaps the most common
retaliatory action by an incumbent faced with a low-cost entrant is a price reduction
(Goolsbee and Syverson 2008; Guiltinan and Gundlach 1996; Simon 2005). One
motivation for this action, which might not be toward profit-maximizing in the short run,
is to increase sales volume and inhibit the challenger from gaining a minimum efficient
scale – increasing the challenger’s cost of production and cutting profit margins
(Steenkamp et al. 2005). By reducing prices, incumbents send a clear signal to the
challenger: that they have low enough production costs which enable them to compete
aggressively on prices. This scenario anticipates that market becomes less attractive for the
newcomers since they will achieve lower sales because of the incumbent’s aggressive price
cuts (Hendricks and McAfee 2006).
Dropping prices may also signal something more subtle, yet even more powerful to
the challenger: that a particular market is worth defending. Since the newly-arrived
challenger has asymmetric (i.e., less) information about the market, in particular about its
future value, this signal is particularly informative to adjust expectations about market
profitability and opportunities in the long run (Hsieh, Tsai, and Chen 2015; Porter 1985).
From the challenger’s standpoint, rational incumbents defend the market by sacrificing
short-term profits in hopes of recouping that loss in the long run (Guiltinan and Gundlach
1996; Porter 1980). In sum, the challenger encountering lower incumbents’ prices faces
mixed signals with respect to market attractiveness. On the one hand, incumbents’ price
cuts might indicate their competitive advantage due to the low production cost, thus, signal
9
the new entrant that the market won’t be as attractive as it had expected before the entry;
On the other hand, the challenger might interpret incumbents’ price-cuts as signals that
there is a strong market opportunity that incumbents consider worth protecting, thus the
market could be really attractive.
We argue that the new entrant is more likely to view the incumbent’s price-cut as a
bluff (Prabhu and Stewart 2001; Sunny Yang and Liu 2015) because it suspects the
incumbent ability to sustain the profit loss for a long term. Since the price cut is an easily
reversible action, the new entrant expects the incumbent to revert the price soon (Hambrick,
Cho, and Chen 1996). Thus, we expect that the challenger is more likely to evaluate
incumbents’ price cuts as the ‘market opportunity’ signal, thereby, will postpone its exit
hoping to gather more accurate information about true market profitability and the
incumbent’s intention in the future (Hitsch 2006).
However, if the incumbents’ lower prices persist over a much longer horizon, we
predict that the low production cost signal (i.e., low demand for the newcomer’s services)
will be stronger and more credible. In this scenario, the challenger’s uncertainty about the
incumbent’s ability to sustain low prices and its intention to do so diminishes, and it
becomes increasingly clear that the incumbents are not bluffing and are instead committed
to fiercely defending their market for a long time, making market less attractive (Chen,
Kuo-Hsien, and Tsai 2007; Prabhu and Stewart 2001). Thus, the challenger will have little
doubt it is time to leave and prevent further losses. Given all the above, we hypothesize
that:
H1: The effect of incumbents’ post-entry price cuts on the challenger’s time-to-exit
is at first positive and becomes negative later on.
10
Incumbents’ service tactics
Besides price-cut, an incumbent can also invest in two well established dimensions of its
service strategies, namely service convenience and service quality (Andreassen, van Oest,
and Lervik-Olsen 2018; Colwell et al. 2008; Farquhar and Rowley 2009; García-Fernández
et al. 2018; Parasuraman, Zeithaml and Berry 1988; Seiders et al. 2007; Thuy 2011). The
perceived value of service is a result of what the consumer sacrifices (negative dimension)
and gains (positive dimension) in return. The positive dimension indicates some benefit
that consumer receives from service, such as quality (Parasuraman and Grewal 2000;
Zeithaml 1988). The negative dimension, which reflects non-monetary expenses that
consumer incurs such as time and effort to consume the service, referred to as service
convenience. High service convenience reduces non-monetary costs such as the time and
effort to receive and consume the service (Collier and Kimes 2013; Colwell et al. 2008;
Farquhar and Rowley 2009; Zeithaml et al. 2006).
Incumbents’ service convenience and the challenger’s exit timing. As mentioned
earlier, service convenience is a consumer’s perception of time and effort spent buying or
using a service (Berry, Seiders, and Grewal 2002). High convenience improves customer
satisfaction, increases switching costs and enhances purchase and repurchase likelihood
(Rust, Lemon, and Zeithaml 2004; Seiders et al. 2007; Voss, Godfrey, and Seiders 2010).
Prior literature has shown that service convenience can be improved in more ways than
one (see Berry, Seiders, and Grewal 2002 and Seiders et al. 2007 for a review of
convenience types). For example, firms can offer better access to their services by making
them available longer and in new and more convenient locations, more days with longer
operating hours (Collier and Sherrell 2010). This access convenience is particularly salient
11
in the case of non-separable services where customers must be present at the time of service
delivery and consumption. Improving service convenience is a less reversible action in the
short run because it can require a substantial investment and other commitments. For
example, improving access convenience by opening more stores in easily accessible areas
could cost millions of dollars. Thus, a challenger faced with incumbents’ service
convenience improvements is more likely to interpret them as a credible signal of the
incumbents’ commitment to defend a market and believes that the incumbent will “stick to
it action” i.e., service convenience improvement. More formally, we propose that:
H2: The effect of incumbents’ post-entry service convenience improvements on the
challenger’s time-to-exit is negative.
Incumbents’ service quality and the challenger’s exit timing. Service quality
which is defined as a gap between customers perceived and expected service (Sivakumar,
Li, and Dong 2014) may also act as a deterrent and influence new entrants (Hauser and
Shugan 2008), regardless of whether incumbents intentionally adjust their service quality
in response to a new low-cost entrant (Bendinelli, Bettini and Oliveira 2016; Prince and
Simon 2014). From the new entrant’s perspective, the existing level of service quality
among incumbents is an informative signal that can influence its time-to-exit. This is
because, in general, incumbents’ high-quality services hurt new entrants, particularly low-
cost ones: high quality improves the demand for incumbents’ offerings, increases customer
satisfaction and willingness to pay (Cho 2014), and generates referrals (Falk,
Hammerschmidt, and Schepers 2010; Homburg, Koschate, and Hoyer 2005; Pauwels and
D’Aveni 2016). Furthermore, incumbents’ high-quality services rely on managerial know-
how and capabilities that are hard to imitate (Parasuraman and Grewal 2000) and are
typically a source of sustained competitive advantages (Bharadwaj, Varadarajan, and Fahy
12
1993; Srivastava, Fahey, and Christensen 2001). Moreover, since incumbents’ prior
investments in quality are not easily reversible actions, the challenger sees them as credible
signals of a strong commitment to protect the market (Chen, Smith, and Grimm 1992;
Hambrick, Cho, and Chen 1996), thereby reducing the challenger’s uncertainty about the
market outlook in the near and far future.
Accordingly, we expect that the higher the levels of service quality of incumbents
the less the new low-cost entrants can reap the expected benefits from the market – and are
more likely to exit sooner than later. More formally, we hypothesize:
H3: The effect of incumbents’ post-entry service quality on the challenger’s time-
to-exit is negative.
Our conceptual framework, summarized in Figure 1.1, rests upon two key
assumptions: (1) the new low-cost entrant has expectations but is uncertain about
incumbents’ ex-poste marketing actions (Chen et al. 2002; Montgomery et al. 2005) and
thus uncertain about the market outlook in the post-entry period (Claussen, Essling, and
Peukert 2018; Ethiraj and Zhou 2019) and (2) the incumbents’ activities may, intentionally
or unintentionally, work as informative signals by which the challenger reduces its ex-poste
uncertainty about the market outlook (Luoma et al. 2018; Marcel, Barr, and Duhaime 2011;
Zajac and Bazerman 1991). Although before entering any market, new entrant would
definitely have studied the market to the best possible extent, have assessed market
attractiveness and may expect to confront with the incumbent’s potential responses in terms
of changes in pricing and service strategies, there is still considerable uncertainty with
regards to the incumbent’s post entry activities (Luoma et al. 2018). Prior literature also
corroborates this argument. For example, conducting an experimental study, Montgomery
et al. (2005) found that due to the uncertainty and ambiguity associated with incumbents’
13
future behaviors, managers usually do not (cannot) consider incumbents’ reactions when
making market entry decisions 8. In line with our essentially exploratory positioning, we
estimate the reduced-form relationships between marketing tactics and an entrant’s time to
exit.
EMPIRICAL ANALYSIS
Data Sources and Industry Context. This Airline industry is particularly well suited for
our purposes because each one of the thousands of routes between any two airports is
considered a unique market, where entries and exits are frequent and easily observed, and
the identification of new entrants and existing incumbents is well established (Dixit and
Chintagunta 2007; Ethiraj and Zhou 2019; Prince and Simon 2014). We focus on market
entry and exit by low-cost carriers (LCC)9, which are frequent in this industry (Ethiraj and
Zhou 2019; Prince and Simon 2014).
Our data cover market-level information, carriers’ characteristics, and marketing
activities over time, from the first quarter of 1997 through the fourth quarter of 2016. We
limited our dataset to 11 major airlines that have the most complete data and remained
significant players in the U.S. airline industry throughout that period: five low-cost carriers
8 Our dataset also verifies this assumption as almost 50% of market exit incidences occurred within 1 year
after the entry. This indicates that the new entrants cannot fully assess the market condition before the entry
and are faced with a large uncertainty about market attractiveness when making any entry decision.
Moreover, on average incumbents reacted to the entry by cutting their prices by 5%. However, these reactions
are distributed with high variability ranging from 80% price cut to 92% price rise. This also validates
Montgomery et al. (2005) argument that precise prediction of the incumbents’ post entry reactions is not
viable. 9 Market entry (exit) is an important strategic decision which requires a strong motivation. For the low-cost
carriers, the main factor that encourages them to enter (exit) a new market is the market profitability.
Whereas, for the major airlines there are several factors in play. For example, a route profitability is not the
only factor that affects entry (exit) decision, its contribution to the entire network profitability also matters
and this factor will affect their decision of entry and exit. Moreover, major airlines might have motives other
than profitability when entering a new market (i.e., they want to establish a foothold in the competitor’s’
main turf in order to prevent any further moves by that competitor in their own turf). So, to avoid any
confounding effects, in this research, we are focusing on the low-cost entrants
14
including AirTran, Southwest, JetBlue, Frontier, and Spirit and 6 major airlines including
Delta, American United, Continental, Northwest and, US Airways). Our sample represents
an expansion on the previous studies that just explore Southwest entries (Dixit and
Chintagunta 2007; Ethiraj and Zhou 2019; Prince and Simon 2014). In each route, we work
with only quarterly observations in which a carrier transports at least 500 passengers
between the origin and destination airports (see Dunn 2008 for similar criteria). This
restriction ensures we are dealing with LCCs that have invested a minimum level of
resources to gain market share after entry. Also, to avoid dealing with differences between
major vs. low-cost incumbents, we only use routes where no other LCC incumbents operate
at the time of entry, nor afterward.10
Our dependent variable, time-to-exit, is the time elapsed between a challenger’s
market entry and exit dates and is measured in quarters. Following Dixit and Chintagunta
(2007), we consider that the LCC has exited a market if it has not served the market for
two consecutive quarters. In our empirical analysis, we use 13,057 observations,
comprising eighty quarters and 1,192 market entries by any of the five low-cost carriers,
555 of which ended up in an exit at some point. The empirical distribution of market exits
over time is depicted in Figure 1.2. Market entries that do not end up in an exit by the fourth
quarter of 2016 are considered right-censored observations, which are also dealt with in
the hazard model of time-to-exit.
A Split-Population Hazard Model. we start by noting that some challengers will
probably ‘never’ leave a market they have entered, which in a hazard or survival models
10 When analyzing firms’ decisions to stay in or leave a market, sunk costs, which are typically unavailable
to researchers, may be a confound (Dixit 1989; Elfenbein and Knott 2015; O’Brien and Folta 2009), and one
difficult to control for empirically. In the airline industry, however, sunk costs are negligible (see Cabral and
Ross 2008).
15
are often referred to as ‘cured subjects’ or ‘long-term survivors’ (Klein et al. 2016). In this
situation, where there is a mixture of two subsamples, classical survival models may lead
to a biased estimation (hazard models implicitly assume all cases will, sooner or later,
experience the event of interest).11 To overcome this issue, we use a split-population hazard
model (Bertrand et al. 2017; Prins and Verhoef 2007; Sinha and Chandrashekaran 1992).12
We use a mixture model, consisting of a logistic regression for the proportion of
new entrants that ‘never’ exit the market and a survival regression for those that do (see
Dirick et al. 2017). We specify the logit part of the model as a function of pre-entry average
market conditions (please see Appendix B) because they reflect the type and level of
required resources that determine market survival in general, i.e., irrespective of time
(Helfat and Lieberman 2002; Ito and Lee 2003). We model the hazard rate of a given
challenger in quarter j as a function of the baseline hazard rate, and market- and firm-
specific factors. In line with other studies in marketing and strategy (Geroski, Mata, and
Portugal 2010; Nikolaeva 2007; Risselada, Verhoef, and Bijmolt 2014), we also include in
the hazard regression the incumbents’ marketing activities themselves and their interaction
with time, which enables us to assess whether the effect of incumbents’ marketing-mix
varies over time. Specifically, hi(tj) is specified as:
hi(tj) = h0(tj)exp{β0 + β1IncPostPriceCutij + β2IncPostFreqij + β3IncPostPeakFreqij + β4IncPostOTPij
+ β5[IncPostPriceCutij × f(tj)] + β6[IncPostFreqij × f(tj)] + β7[IncPostPeakFreqij × f(tj)] +
11 It is impossible to know, from observed data, whether a low-cost carrier will never exit a given route or is
just right-censored. In the unlikely case that all carriers would exit, the split-population model would
incorrectly identify some of them as being cured, i.e., never exit (see Jaggia 2011). This is more likely in
short datasets. Because our dataset leaves plenty of time for those carriers that entered routes long time ago
to exit them, we believe that a split-population model is more realistic than a hazard model that assumes the
data are right-censored. 12 While some challengers that remain in the route at the end of the observation period are likely to exit some
time in the future, it is reasonable to assume that some will ‘always’ be immune to incumbents’ marketing-
mix reactions (but may still exit in the far future for other reasons).
16
β8[IncPostOTPij × f(tj)] + β9ChllgPriceij + β10ChllgSizeij + β11MMCij + β12Demandi + β13Hubij +
β14NIncij + β15Distancei + β16FuelPricej + β17ChllgNetworkij + β18IncNetworkij + β192ndEntryi +
β20-28IncChllgi + β29-48Yearj} (1.1)
where f(t) = t + t2 + Ln(t) is a flexible time function (Chandrasekaran and Tellis 2011) and
the right-hand side independent variables are operationalized as described below.
Time-variant Independent Variables in the Hazard Regression
Price-Cut. we compute incumbents’ post-entry quarterly price-cuts, IncPostPriceCutij, as
one minus the weighted average price on the route i in quarter j after entry divided by the
weighted average price over eight pre-entry quarters, where incumbents’ market-shares
serve as weights. The use of weights based on market shares ensures that the relative
competitive strength (leader vs. followers) of incumbents in a market, and their impact on
demand, is preserved (Dixit and Chintagunta 2007), and the use of a ratio accounts for pre-
vs. post-entry differences.13
Service convenience and service quality. Carriers offer consumers more
convenient access to their service by increasing the frequency of flight departures in
general (Berry and Jia 2010; Brueckner, Lee, and Singer 2013) and more flights in peak
times in particular (Huse and Oliveira 2012). These factors affect passengers’ choice of
airline because travelers are both price- and time-sensitive (Shaw 2007). Accordingly, the
first of the incumbents’ service convenience measure in route i in quarter j, IncPostFreqij,
is the average number of non-stop flights in quarter j post-entry divided by that in the pre-
13 The use of market-share weighted averages assumes the low-cost entrant looks at the actions of a
‘representative incumbent’ while still preserving market-share differences. In other words, the actions taken
by say an undisputed market leader will show more strongly than those with negligible market shares. In
such cases, a new entrant is likely to pay more attention to ‘who does what’ rather than second-order effects
such as ‘who did what first and when’.
17
entry stage, using again market-shares as weights and eight pre-entry quarters. we do the
same for IncPostPeakFreqij, the percentage of flights that depart during daily peak time,
i.e., 7-10am or 3-7pm on weekdays (see Oliveira and Huse 2009; Sengupta and Wiggins
2014).
According to the marketing literature, one of the main indicators of service quality
in the airline industry is the percentage of flights that arrive on-time (Grewal,
Chandrashekaran, and Citrin 2010), which is available at route-level (see Prince and Simon
2014). To measure the on-time performance variable IncPostOTPij, we use the market-
share weighted average of the percentage of incumbents’ flights on the route i in quarter j
that arrive on-time.14
In equation (1.1), the derivative of log(hi(tj)) with respect to incumbents’ price cut,
flight frequency, peak-time flight frequency, and on-time-performance, is β1 + β5f(tj), β2 +
β6f(tj), β3 + β7f(tj), and β4 + β8f(tj). If β5, β6, β7, and β8 are non-zero and significant, we find
support for the post-entry time-variant effects of marketing tactics on time-to-exit. The
main effect and the time interaction effect combined determine whether the direction or
sign of the overall effect changes over time. For example, if the estimates for β1 and β5 are
such that β1 + β5f(tj) is positive after entry and then turns negative, there is support for H1,
suggesting that incumbents’ deeper post-entry price cuts lengthen a challenger’s expected
time of exit at first, but they shorten it afterward (see Risselada, Verhoef, and Bijmolt 2014
for a similar interpretation).
14 Since, on average, incumbents’ peak frequency and OTP did not change at the time of a challenger’s entry
(see Figure A.1 in Appendix A), we do not use changes relative to the pre-entry period but only their levels.
We re-estimated our model using peak and OTP reactions and our key findings are robust to these alternative
model specifications.
18
Time-variant Control variables and market network structure. In equation (1.1),
ChllgPriceij is the average one-way fare charged by the low-cost to its passengers on the
route i in quarter j post-entry. ChllgSizeij is the natural log of the number of passengers that
are carried by the challenger in quarter j. Demandi is the geometric mean of the population
in the endpoint cities. NIncij is the total number of incumbents in route i in quarter j, and
Distancei the distance between two endpoint airports for each route.
Since fuel costs are one of the largest expenses for airlines and account for almost
30% of their operating costs,15 we include quarterly FuelPricej in our model. And because
airlines often compete against each other in many markets simultaneously, which
influences their competitive behaviors (Baum and Korn 1996; Jayachandran, Gimeno, and
Varadarajan 1999), we also control for a multimarket variable (MMC). Since there is a
possibility that another low-cost challenger enters a market before the first entrant’s exit,
and this second entry influences the first challenger’s exit timing, we also include a
2ndEntryi variable in the hazard regression ( = 1 if a second low-cost challenger stepped in,
zero otherwise).
In the airline industry, what happens in one market – including who comes in and
who leaves, and when – is not entirely independent from what happens in all other markets,
since the different (geographical) markets are naturally connected by the very nature of
routes linking any two airports, and some airports are more central than others. To account
for this interdependency of the different markets, we control for and include in our
econometric model a challenger and incumbents’ route importance or route centrality
15 https://www.iata.org/pressroom/facts_figures/fact_sheets/Documents/fact-sheet-fuel.pdf
19
within an LCC’s network, ChllgNetworkij and IncNetworkij, respectively (Please see the
Appendix D for a discussion of the operationalization of MMC and the route importance).
Finally, we include a set of yearly dummies Yearj to capture unobserved time-
varying macroeconomic factors such as shifts in demand and costs of production, and other
unobserved time factors (Greenfield 2014). In addition to the incumbent- and challenger-
specific covariates, we also account for any unobserved firm-specific heterogeneity by
implementing a fixed-effect model and include a set of challenger and incumbent dummies,
IncChllgi, to capture potential unobserved incumbent- and new entrant-specific factors.
Table 1.2 lists all control variables and how we operationalize them.
Following common choices in cure models (see Jaggia 2011), we use a Weibull
distribution in the baseline hazard function and a log(-log) link function in the incidence
part. We estimate the model parameters in Stata using the command cureregr (which
uses maximum likelihood estimation). We use route-level clustered standard errors that
make our hypotheses testing more conservative and enable us to control for unobserved
route-specific factors that might influence a challenger’s time-to-exit (Eilert et al. 2017;
Panagopoulos et al. 2018).
Endogeneity
Before we mitigate concerns about the endogeneity of incumbents’ prices to the hazard
function of the low-cost carrier (in which case the estimated price effect may be biased and
inconsistent), we note the following. First, unobserved demand shock is not the primary
driver of the results. If the demand drives the incumbent’s price and the challenger’s exit
decision simultaneously, we should see incumbents dropped the prices less when the
entrant’s exit likelihood is low, not the other way around. Second, it is unlikely that
20
incumbent carriers set prices based on a newcomer’s likelihood of exiting the market at a
particular quarter (for a detailed discussion, please see Dixit and Chintagunta 2007, page
162). We argue that if the incumbents set prices based on the entrant’s risk of staying-
in/exiting the market (rising prices when the exit likelihood is high), they should follow
the same line of reason for their flight frequency (setting low flight frequency when the
exit likelihood is high), however, we observed that incumbents follow two different paths
with regards to price and frequency. Third, and although “there is no direct evidence from
the firm side (for example, from pricing experiments) that endogeneity biases are large in
panel or time-series data (Rossi 2014, p. 670),” we explicitly control for several demand
factors common to both prices and newcomers’ time-to-exit in our model, namely route
demand, ingredient costs (i.e., fuel prices), and competition information (number of
incumbents in the market per quarter). Admittedly, other unobservable demand factors can
be thought of but it is hard to imagine that those would have a larger impact on the
dependent variable and would drive a larger portion of the variation in incumbents’ prices
than the ones we observe and do include in the model (see Rossi 2014). Finally, we include
route-, time-, incumbent-, and challenger-specific fixed effects that capture unobserved
factors at these levels and will alleviate the endogeneity due to the omitted variables (Ebbes
et al. 2016; Ketokivi and McIntosh 2017; Rossi 2014).
Although before addressing an endogeneity, a strong and convincing argument
must be made that there is first order endogeneity problem (Ketokivi and McIntosh 2017;
Rossi 2014) - which we believe is not - still, to empirically explore whether price
endogeneity is a major concern in the context of our nonlinear hazard model and investigate
the robustness of our findings more formally, we follow Risselada, Verhoef, and Bijmolt
21
(2014) and Terza, Basu, and Rathouz (2008) approaches. We implement a two-stage
residual inclusion estimation method (2SRI) using instrumental variables, which is an
extension of the popular two-stage least squares (2SLS). Our analysis suggests that price
is not endogenous (Please see Appendix C for the comprehensive discussion of 2SRI
method). In this situation, Ebbes et al. (2016) recommend that results that come from a
regression without the instrument should be used for inference. Accordingly, in the next
section, we report the findings from our initial model specification, treating price as an
exogenous factor.16
RESULTS
Table 1.3 presents the results of the split-population hazard model that estimates the impact
of incumbents’ marketing tactics on, simultaneously, the challenger post-entry exit
likelihood and the challenger’s time-to-exit. The fit of the model is significantly better than
one with no marketing variables (χ2(16) = 337.09, p < .01) and better than a model without
a flexible polynomial time function (χ2(12) = 349.68, p < .01). Notice that the model is
parameterized in such a way that a positive coefficient in the logit or incidence regression
implies a positive effect on the challenger’s exit likelihood, while a positive coefficient in
the hazard or latency regression implies a positive effect on the hazard rate, i.e., a negative
effect on exit timing, as the expected time for a market exit is shortened. We first present
briefly the results in the exit likelihood part of the model and then turn to the results in the
exit timing, which is our main focus. In the latter, we are particularly interested in knowing
whether time moderates the effect of incumbents’ marketing-mix – in terms of prices and
16 We also tested for the endogeneity of flight frequency using “ConnPass” as an IV. We followed the same
procedure and found that the p-value associated with the residual coefficient was not significant (p > .1)
indicating that endogeneity of service convenience is also not a big concern.
22
service – on a new entrant’s exit timing (see Figure 1.1) and, if it does, in the way we
predicted.
Exit likelihood. The results from the logit part of the model reveal that the higher
the challenger’s quarterly prices, the lower the exit likelihood (γChallgrPrice = -.00860, p <
.05), which may be seen as a sign that the market is financially attractive. The overall exit
likelihood of a low-cost challenger is also significantly affected by route pre-entry
marketing environment. Specifically, the higher the incumbents’ pre-entry prices, the
lower the low-cost challenger’s exit likelihood (γIncPrePrice = -.00936, p < .05), possibly
because, at the time of entry, the new entrant’s low-cost proposition was a particularly
compelling one among price-sensitive consumers that higher priced mainstream carriers
were not serving effectively. The effects of incumbents’ pre-entry service are mixed,
however. Low-cost challengers were less likely to leave a market where incumbents were
offering a higher flight frequency at the time of entry (γIncPreFreq = -.00327, p < .05), which
suggests the market was underserved, yet they were more likely to leave markets where
incumbents were using larger aircraft at the time of entry (γIncPrePlaneSize = 9.20400, p < .01),
a level of quality that new low-cost entrants were perhaps not ready to compete with.
Exit timing (or time-to-exit). we start by describing the results regarding the effects
of control variables that may be confounded with the effect of incumbents’ marketing
tactics on the exit timing of a low-cost entrant. As indicated in Table 1.3, control variables
are measured at route-, challenger- and network-levels and some are time-variant (e.g.,
number of incumbents and fuel price).
Control variables. All control variables but one (whether there is an incumbent’s
hub in one of the two endpoint cities; p > .10) are highly significant explaining a new
23
entrant’s exit timing. We briefly discuss these results. A challenger’s price has a negative
and significant effect on the hazard rate, i.e. it increases the expected timing of exit
(βChallgrPrice = -.00137, p < .01), in line with its effect on exit likelihood irrespective of time,
as described before. A challenger’s size, however, has the opposite effect (i.e., a positive
significant effect on the hazard rate): larger challengers tend to exit sooner (βChallgrSize =
.00009, p < .01), perhaps an indication of ‘too heavy a load’. All market-level
characteristics – whether there has been a second challenger entering the market (β2ndEntry
= .22832, p < .01), the larger the distances traveled (βDistance = .00965, p < .01); a larger
number of incumbents (βNInc = .10274, p < .01) and of other markets where the challenger
faces the competition of the same incumbents (βMMC = .36974, p < .01); and higher fuel
prices – significantly shorten the exit timing. These effects could be expected from an
economic point of view. For instance, the cost efficiency of low-cost challengers compared
to that of mainstream incumbents shows up more strongly on shorter travel distances as
longer routes become too costly to serve (Joskow, Werden, and Johnson 1994). Not
surprisingly, the exception is market demand, which decreases the hazard rate, i.e.,
lengthens the exit timing of the new low-cost entrant (βDemand = -.20187, p < .01). Similarly,
the importance or centrality of a route within the challenger’s network has a significant and
negative effect on the hazard rate (βChllgNetwork = -1.56558, p < .01), meaning the expected
time to exit is longer. Conversely, the more the route is important to the incumbents, the
sooner the challenger’s exit time (βIncNetwork = .00275, p < .01).
Incumbent’s price cuts, service convenience, and service quality. As reported in
Table 1.3, incumbents’ post-entry marketing elements have a significant effect on a new
entrant’s hazard rate and, consequently, on its exit timing. While service convenience, i.e.,
24
flight frequency during regular-time (βIncPostFreq = .32821, p < .01; but not during peak-time,
p > .10) has a positive effect on the hazard rate, i.e., it shortens the entrant’s expected time
to exit, both price cuts (βIncPostPriceCut = -1.43601, p < .01) and service quality (on-time
performance; βIncPostOTP = -1.51105, p < .01) negatively affect the hazard rate, i.e., they
lengthen the entrant’s expected time to exit. These main effects are only part of a larger
story, however. As our results reveal, the passage of time has a significant moderating
effect on the relationship between incumbents’ post-entry marketing elements and a new
low-cost entrant time-to-exit.
In Figure 1.3, we plot the overtime effects of the incumbents’ tactics on the
challenger’s TTE with corresponding 95% confidence intervals. The overall effects of
incumbents’ marketing elements on a challenger’s exit timing as time goes by take different
shapes. The effects of incumbents’ post-entry price-cuts have a U-shape over time (top-left
of Figure 1.3), as they first lengthen the challenger’s expected time to exit until roughly
quarter 10 (i.e., estimated overall effect on the challenger’s exit timing is positive though
decreasing) and then shorten it almost until the end (i.e., estimated overall effect on exit
timing is negative though increasing), which supports H1.
The effect of incumbents’ post-entry flight frequency has somewhat of an S-shape
over time (bottom-left of Figure 1.3). At first, and until roughly quarter 10, it shortens the
challenger’s expected time to exit. Afterward, and until approximately quarter 50, its 95%
CIs include zero, i.e., the estimated overall effect on exit timing is not significant. It then
lengthens the expected exit time until the end of the observation period. This result lends
only partial support to H2.
25
The effect of incumbents’ post-entry peak-time flight frequency has an inverted U-
shape (top-right of Figure 1.3): it is non-significant at first (the estimates include zero
within the 95% CI); it lengthens the expected time to exit until about quarter 55, and it
becomes insignificant again afterward. Perhaps increasing peak-time frequency is a sign of
incumbents’ strengthening their core positioning among business customers (Kumar 2006;
Wang and Shaver 2014), which is not the typical target market of low-cost carriers.
Competition is thus less intense, and the challenger has a higher chance of survival.
The effect of incumbents’ post-entry flight on-time performance (i.e., service
quality) on exit timing is monotonically decreasing over time (bottom-right of Figure 1.3);
It is increasingly negative, i.e., it increasingly shortens the expected time to exit after
quarter 5, before which it has the opposite effect (i.e., the estimated overall effect on exit
timing is positive though decreasing). As we mentioned earlier, the initial stage after the
entry is characterized by high uncertainty and challenger will gather and interpret any
market signal to reduce its uncertainty regarding the new market. Hsieh et al. (2015)
indicate that firms usually consider competitors as external reference points and use any
signal from them (both intended and unintended signals) to justify future decisions. One of
the main concerns of LCCs is to keep the turnaround time as low as possible (Berry and
Jia 2010) because it enables them to reduce its cost per each seat-mile. In the case of high
delay, the challenger cannot benefit from this advantage and serving that market will be
costly. Since a big portion of delays might be due to the other airport-level factors that are
out of the airline’s control, when large and established incumbents perform poorly (i.e.,
high delay) in the market despite having more resources and capabilities, managers of a
challenger firm may infer that the conditions in the new market is unduly challenging,
26
therefore, they may not have a sufficient resource and capability to contest in the new
market and have a lower chance to achieve their desired goals. In this situation, “larger
competitors’ negative performance will create an unfavorable expectation for return on
commitment,” (Hsieh et al. 2015, p.43) therefore, further resource commitment is no longer
justifiable, and a new entrant might decide to exit the market. In sum, our results lend
support for H1, partial support for H2, and support for H3 in about quarter 5 and beyond.
We discuss the implications of our findings next.
Robustness Checks
Market definition in the airline industry. Although several studies using airline data have
defined a market as a route between two airports, prior literature has questioned this
definition when one of the endpoints is a large metropolitan area with multiple commercial
airports (Brueckner, Lee, and Singer 2014). The issue is whether these multiple airports
are representing a single destination for passengers, or each of them should be considered
as a separate destination (Brueckner, Lee, and Singer 2014). To explore how market
definition (city-pairs vs. airport-pairs) may affect our findings, we treat multiple airports
in large cities as a single destination (origin) by grouping them as suggested in Brueckner,
Lee, and Singer (2014). For instance, the routes from the three airports in New York
(Newark, John F. Kennedy, and La Guardia) to Atlanta, were grouped as a single route,
New York-Atlanta. We re-analyzed our model using a new set of market entry and exit
observations and find that our key findings are not sensitive to city-pairs vs. airport-pairs
market definition (see Table 1.4-column 1).
Flight OTP specifications. Also, to test the sensitivity of our results to the
definition of flight delay, we re-estimate our model using two alternative measures
27
suggested by the US Department of Transportation (DOT) and used in previous studies: an
arrival at destination 15 and 30 minutes late (as the proportion of incumbents’ flights on
route i in quarter j that arrive that late; see Prince and Simon 2014). The results suggest
that our key findings are not driven by the definition of delay (Table 1.4 - columns 2 and
3).
Southwest and AirTran merger. In 2011, Southwest acquired AirTran, the second
largest LCC in the U.S. airline industry. From that point on, such major network
restructuring might have influenced Southwest and AirTran time-to-exit decisions –
something we should avoid being confounded with our focal marketing tactics. Thus, to
rule out an alternative explanation due to this event, we conducted our analysis using a
subsample that excludes all exit events that occurred after 2011. To make sure our results
are robust to the cut-off year, we also re-estimated our model on other subsamples using
2010 and 2012 as cut-off points. The results indicate that our key findings are robust to the
LCC merger in the U.S. airline industry (Table 1.4-column 4).
Southwest as a low-cost or as a major carrier. Although Southwest was originally,
and in our observation period, a low-cost carrier, it grew significantly and became the
number one carrier in the US in terms of number of domestic passengers (Dixit and
Chintagunta 2007). Thus, one might argue that Southwest is no longer a low-cost carrier,
and it is more like a major carrier that might behave differently from other low-cost carriers,
and the factors that affect its survival may be different. Following Dixit and Chintagunta
(2007), we also analyzed the data without Southwest entry-exit observations. Since the
effects of key covariates are similar, we present the results considering Southwest as a low-
cost carrier (Table 1.4-column 5).
28
Challenger’s post-entry marketing strategies. Low-cost carriers have been
reporting their flight fares to DOT since 1990; however, they started reporting OTP and
flight frequency data at different points in time during our observation period. Thus, these
variables were missing during the post-entry period for more than 60% of route-challenger
observations. Given this limitation, in our model, we just controlled for a challenger’s
price. However, as a robustness check, we re-analyzed our model on a subsample of routes
where flight frequency and OTP data were available for challengers in the entire post-entry
period. Our key findings are robust to this model specification – and the results of this
additional analysis indicate that a challenger’s higher flight frequency reduces its exit
likelihood, whereas OTP and peak frequency do not significantly affect its time-to-exit
(Table 1.4-column 6).
GENERAL DISCUSSION
Prior research has shown that incumbents usually react to a rival’s market entry by
adjusting their marketing tactics (Goolsbee and Syverson 2008; Luoma, et al. 2018;
Shankar 1997). However, there is little evidence regarding how these adjustments affect
the entrants’ post-entry exit. Drawing from the related notion of action irreversibility and
information economics, the primary purpose of this study was to examine the link between
incumbents’ marketing tactics and a challenger’s exit likelihood, over time and at the
market level. Next, we summarize our main findings and contributions.
Research Contributions
From a theoretical perspective, by examining the link between a challenger’s time-to-exit
and incumbents’ marketing-mix, our research offers new insights into the market exit
literature and addresses calls of prior researchers for an investigation into other factors that
29
might influence market survival (Dixit and Chintagunta 2007). Our findings suggest that
incumbents’ marketing tactics related to price, service quality, and service convenience
impact a challenger’s exit timing, and that the time elapsed after entry works as a moderator
of those effects.
Specifically, the results of this study indicate that while incumbents’ price cuts
increase the challenger’s exit likelihood later after entry, they reduce the exit likelihood
early after entry. On the other hand, incumbents might be better off by not investing in their
quality immediately after a low-cost carrier has challenged their market, because our results
indicate that the lower their levels of service quality, the higher is the challenger’s exit
likelihood early after entry. This finding, though seemingly counterintuitive, may help
explain why incumbent airlines have been seen to not improve quality in response to the
entry of a low-cost carrier (Prince and Simon 2014). We believe the deeper study of this
phenomenon in future research is warranted. Moreover, the findings of our research
indicate that investments of incumbents in service convenience increase the challenger’s
exit likelihood early after entry.
Previous studies in marketing have highlighted the importance of service
convenience and noted that empirical research should pay more attention to the concept of
service convenience as a construct in its own right (Collier and Sherrell 2010; Farquhar
and Rowley 2009; Rust, Lemon, and Zeithaml 2004). By drawing a distinction between
service quality and service convenience, our study is among the first to empirically
investigate the link between service convenience and a new entrant’s exit decision. As
such, our findings contribute to the service convenience literature (Collier and Sherrell
2010; Farquhar and Rowley 2009; Obeng, et al. 2016; Seiders et al. 2007) by recognizing
30
service investment and in particular convenience investments as a strategic weapon at a
firm’s disposal that can be effectively employed in a competitive environment to protect
markets against a rival’s attack.
The findings of our work may also suggest a new rationale for why a delayed
reaction might be an optimal strategy for incumbents. Several empirical studies have found
that incumbents sometimes delay their reactions to a challenger’s market entry and
underscore firm inertia, lack of managerial capability, capacity limitation and so on as
factors that cause this delayed response (Bowman and Gatignon 1995; Robinson 1988).
Kalra, Rajiv, and Srinivasan (1998), however, have proposed that incumbents’ immediate
reactions in the form of price cuts are an implicit acknowledgment of the entrant’s high
quality and enhance the attractiveness of the challenger’s product to customers. Similarly,
we suggest that incumbents’ immediate reactions might send mixed signals with regards
to market attractiveness, thereby increasing the new entrant’s uncertainty about the market
condition. In this situation, the challenger is likely to delay an exit decision until more
accurate information is gathered and hence incumbents may be better off not reacting or at
least delaying their pricing responses to the entry.
Moreover, Luoma et al. (2018) indicate that since it might be hard to justify the
subsequent price increase after the new entrant’s exit, aggressive price reaction to the entry
might have a persistent negative impact on the incumbent’s profitability. Drawing on firm
managers’ competitive reasoning, our study provides a novel firm-level reason for why
immediate aggressive price cut might not be an optimal action in response to the low-cost
entrant.
31
From a methodological perspective, we applied the cure model to study market
survival throughout the post-entry stages. Unlike typical survival models, the cure model
does not assume that the survival function goes to zero as time goes to infinity, i.e., it does
not assume that all subjects will eventually experience the event of interest. Accordingly,
in our research context several firms probably do not leave the market they have entered
and continue to serve it for a very long time. While we account for a proportion of
challengers that do not leave the market, the cure model enables us to simultaneously
explore the factors that impact the probability of exit and those that impact the timing of
the exit.
Finally, because airlines operate over a network – i.e., their markets are connected
their exit decisions in one market may depend on and influence the exit decisions in another
market. In other words, the importance of each market (i.e., route) is evaluated not only by
its stand-alone profitability but also by the passenger-flow contribution that it brings to the
carrier’s total carried passengers (Boguslaski, Ito, and Lee 2004; Dunn 2008). Thus, we
included in our econometric model measures that described network structure and assessed
their impacts on the firm’s survival. We find that the higher the route importance within
the challenger’s network, the less likely it is that a challenger will leave that market.
However, the higher the route centrality within the incumbent’s network, the higher would
be a challenger’s exit likelihood.
Managerial and Policy Implications
Our study has implications for both managers and policymakers. When and how to allocate
limited resources to defend markets under attack has long been a vital question for
marketing managers. This research suggests that managers should choose carefully the
32
type, the timing, and the intensity of their defensive responses to entry and offer valuable
insights for practitioners to efficiently assign marketing expenditures. More specifically,
our findings delineate that aggressive price-cut although is the easiest and fastest way of
response, may not be the most efficient strategy to repel the new entrant in the short run.
By cutting prices, incumbents intentionally signal their production efficiency to the
challenger to make the market less attractive. However, since the price-cut represents less
credible and unsustainable strategy, the challenger may interpret this action as an
opportunity in a market worth defending. So, in this situation, the challenger receives a
mixed signal that increases uncertainty. As such, a newcomer is prone to stay longer and
let additional time go by to gain more information from the market (i.e., encouraged to
‘wait and see’); the passage of time reduces uncertainty, enabling the entrant firm to make
a better prediction about the future of the market (Haenlein, Kaplan, and Schoder 2006;
O’Brien and Folta 2009). Thus, we recommend managers to avoid deep immediate price-
cut in response to the entry and advocate for implementing service-based strategies along
with a low to moderate price cut to repel a low-cost entrant.
The findings of this research indicate that incumbents’ post-entry strategies are vital
determinants of the challenger’s survival and suggest that the new entrant will be much
better off if it anticipates the incumbents’ actions in response to the entry. However,
marketing scholars argue that managers usually cannot accurately predict incumbents’
activities and there is uncertainty and ambiguity associated with incumbents’ reactions to
the market entry (Chen et al. 2002; Montgomery, Moore, and Urbany 2005). Our research
provides managers with a better tool to identify markets with a higher chance of survival
regardless of how incumbents react to market entry. For instance, for firms entering new
33
markets, our findings suggest that pre-entry market environment (i.e., history of
incumbents’ prior strategies and available resources) is an important factor that might
affect the challenger’s survival and must be investigated carefully before making any entry
decision. Also, from the perspective of an airline entering a new market, the findings
suggest that potential new entrants should not be deterred by incumbents offering a higher
service convenience if other market factors look favorable.
This study might also provide valuable insights for policymakers. One of the main
roles of policymakers is to promote a fair competitive environment for the benefit of
consumers. For instance, antitrust laws prevent anti-competitive strategies and protect
firms in the case of predatory behavior in response to market entries. Marketing scholars
define predatory pricing as an incumbent’s deliberate price cut, usually below cost or at an
unprofitable level, to squeeze a challenger out of the market (Guiltinan and Gundlach
1996). Our findings reveal, however, that cutting prices in response to entry does not
reduce competition, at least not immediately. But price-cut strategies might still be a
concern for policymakers if they persist long after entry – as we showed, at that time they
do drive challengers out of the market.
Limitations and Future Research
While this study provides novel insights into firm survival, it also faces limitations that
open the way to future research. The fact that the study is limited to the airline industry
implies that the results may apply in another industry somewhat differently. However,
using data from a single industry allows us to eliminate any confounding effects from
extraneous industry-specific factors, thereby improving internal validity (Eilert et al.
2017).
34
Furthermore, although we explored how the type and the timing of incumbents’
marketing activities help them protect their markets, an interesting opportunity for future
research lies in examining the long-term and indirect effects of the incumbents’ marketing
efforts. Clark and Montgomery (1998) indicate that an incumbent’s willingness and ability
to defend its market enhance its reputation as a “credible defender,” and this reputation will
deter potential entrants from attacking incumbents’ markets in the future. It would,
therefore, be important to empirically investigate the long-term and indirect effect of
incumbents’ defensive actions on their performance. In other words, to what extent do
incumbents’ marketing actions in the face of entry deter potential entrants from entering in
the future? Understanding the answers to these questions is important for both managers
and policy-makers.
In addition, we defined a market exit as a complete withdrawal from the market
(operationalized as a binary variable). However, instead of complete withdrawal,
challengers might decide on major downscaling of participation (i.e., reducing the number
of seats available to the customer or flight frequency) while remaining in the market
(Boeker et al. 1997). It would be useful to include information about the level of
participation in a particular route and investigate how the incumbents’ activities affect the
challenger’s service scale. Doing so would give us a better understanding of the difference
between a complete exit from the market and a significant change in the level of
participation in that market.
Marketing and strategy literature classifies the post-entry period into three distinct
stages: (1) an immediately after entry (retaliation or entry) stage; (2) an intermediate
sequencing stage; and (3) a long-after-entry (competition or post-entry) stage (Gatignon,
35
Anderson, and Helsen 1989; Guiltinan and Gundlach 1996; Homburg et al. 2013;
Porter1985). Since each stage has certain characteristics, a challenger’s vulnerability to
the incumbents’ actions might vary over these three stages. Thus, another promising
avenue for future research is to empirically identify these three stages and investigate how
the effect of incumbents’ marketing-mix on a new entrant’s time-to-exit varies across these
stages.
Moreover, following prior studies in the airline industry, we used ‘on-time
performance’ (OTP) as a measure of service quality and both regular and peak time flight
frequency as measures of service convenience. However, incumbents could improve other
aspects of service quality such as mishandling baggage, legroom, and in-flight amenities.
An operationalization of service quality that includes other measures would advance our
current state of knowledge on the effects of incumbents’ service quality on a new entrant’s
exit likelihood.
Furthermore, understanding how loyalty programs could influence incumbents’
marketing activities effectiveness over the post-entry period is another valuable direction
for future research inquiry. For example, if an incumbent possesses a valuable and strong
loyalty program, a price drop or an improvement in service would attract more customers
to the program. Exploring this question will shed more light on the indirect effect of loyalty
programs on firm performance through its impact on competitor’s behavior.
Finally, we acknowledge that our theorizing would suggest a policy-invariant
structural economic model with sequential decisions made under uncertainty capturing a
strategic market environment where the beliefs incumbents and entrants about each other’s
actions matter. Future research could develop and test more flexible dynamic structural
36
models addressing underlying sequential decision-making process. In this study, we took
a step in that direction and hope our findings stimulate further interest in the study of the
market exit phenomenon as a dynamic process involving time-dependent interactions
between incumbents and new entrants.
37
FIGURE 1.1: Conceptual Framework
INCUMBENTS’
POST-ENTRY MARKETING
Controls Market Characteristics
Entrant’s Characteristics
Incumbents’ Characteristics
Network Characteristics
Pre-entry Environment
TIME
PRICE
SERVICE
CONVENIENCE
SERVICE
QUALITY
ENTRANT’S
TIME-TO-EXIT
38
Pro
port
ion
of
Mark
et
Exit
s
Quarters After Market Entry
FIGURE 1.2: Distribution of Market Exits over Time
39
Notes: Hazard regression estimates were multiplied by -1 to depict the effect on exit timing (in the vertical axis). Post-entry quarters are
depicted in the horizontal axis. The solid line represents the average estimated effect, dashed lines represent the 95% confidence intervals.
The Stata lincom command was used to generate mean effects and confidence intervals for each quarter as specified in Equation (1.1),
i.e. using the sum of marketing main effects with their interaction with a flexible polynomial time function (t + t2 + Ln(t)). Stata uses the
variance-covariance matrix to estimate the standard errors associated with these quarterly overall marketing effects.
FIGURE 1.3: Time-Dependent Effects of Incumbents Marketing Tactics on a Challenger’s Exit Timing
-1
-0.5
0
0.5
1
1.5
2
-10 10 30 50 70
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0 10 20 30 40 50 60 70
Service Convenience
(Flight Frequency)
-3
-2
-1
0
1
2
0 10 20 30 40 50 60 70
Service Quality
(OPT)
-1.5
-1
-0.5
0
0.5
1
1.5
0 10 20 30 40 50 60 70
Price-cut Service Convenience
(Peak-time Flight Frequency)
40
TABLE 1.1: Literature Review of Antecedents of Market Exit and Study Contribution
Study Type of
characteristics
Marketing-
mix
Time-
variant
effects
Method Key Learnings
Srinivasan, Lilien, and Rangaswamy
(2004) Firm, product No No
Accelerated
failure time
(AFT) model
Network externalities negatively affect the
survival duration of pioneer entrants.
Min, Kalwani, and Robinson (2006)
Entry timing,
product
innovativeness
No No
Multivariate
hazard rate
analysis
First-mover with a ‘really’ new product
has a high failure rate. Whereas the first
mover that introduces an incremental
innovation can enjoy higher survival
likelihood.
Bayus and Agarwal (2007)
Pre-entry
experience,
entry timing,
product
technology
No No
Discrete-time
hazard
(DTH) model
Introducing products with the newest
available technology increases survival
likelihood. Entrant’s pre-entry experience
and entry timing moderate the link.
Bercovitz and Mitchell (2007) Firm, product No No AFT model
Business profitability, scale, and scope
(product line breadth) during a baseline
period contribute to long-term business
survival.
Dixit and Chintagunta (2007) Firm, market
(size, demand)
Pricing
strategy No
Bayesian
learning
(belief-
updating)
While challenger’s price affects its exit
decision, incumbent’s price is not a
significant driver.
Nikolaeva (2007)
Industry, firm,
product, macro
environment
No Yes DTH model
Publicly traded firms and digital products
increase survival in the beginning, but not
sustainable. Inverted-U between exit rate
and age. Survival decreases with
competitive density and market growth at
41
a time of entry increases with economic
growth.
Johnson and Tellis (2008)
Entry mode,
entry timing,
firm size
No No Multiple
regression
Smaller firms more successful in entering
emerging markets. The entry that involves
high levels of control (e.g., owned
subsidiaries) more successful than low
levels (e.g., licensing).
Franco, Sarkar, Agarwal, and
Echambadi (2009)
Entry timing,
product
technology
strategies
No No Hazard
model
Early entry is beneficial only for pioneers
that are technically strong. However,
pioneers that are low on technological
capabilities suffer from poor survival
rates.
Geroski, Mata, and Portugal (2010)
Firm, market,
macro
environment
No Yes
Semiparamet
ric hazard
model
Larger firms survive longer and this effect
is ‘almost permanent.’ Effect of
concentration at the time of entry has a
strong negative effect on survival.
However, the effect disappears
immediately after entry. Impact of initial
human capital seems to be permanent too.
Wang, Chen, and Xie (2010) Order of entry,
market, product No No AFT model
Pioneers are likely to enjoy a survival
advantage when their product is cross-
generation compatible but within-
generation incompatible.
Homburg, Fürst, Ehrmann, and
Scheinker (2013) Market, product No No
DT-SIR
epidemic
modela
The success of incumbent’s investments
aimed at squeezing entrants out of the
market depends on the length of the
product life cycle (PLC).
Pe’er, Vertinsky, and Keil (2014) Firm, market No No
Cox
proportional
hazard model
U-shaped relationship between new
entrant’s growth rate and the likelihood of
failure, moderated by environment
characteristics.
42
Chadwick, Guthrie, and Xing (2016) Firm No No DTH model
Presence of an HR executive on firms’
TMTs at the time of entry is related to the
firm survival
This study Firm, market
Price,
service
quality,
service
convenience
Yes
Split
population
hazard
model (Cure
Model)
Incumbents’ price-cuts delay
newcomer’s time-to-exit first, speed it
up afterward. Incumbents’ service
convenience speeds up newcomer’s exit
time first delay it afterward. aDT-SIR: discrete-time susceptible-infected-recovered.
43
TABLE 1.2: Variable Operationalization
Pre
-en
try
Ma
rket
ing
Va
ria
ble
s IncPrePricei
The average price over 8 pre-entry quarters and across all
incumbents in route i.
IncPreFreqi Average flight frequency over 8 pre-entry quarters and across
all incumbents in route i.
IncPrePeakFreqi The average percentage of flights during peak hours over 8
pre-entry quarters and across all incumbents in route i.
IncPreOTPi The average percentage of On-time flights over 8 pre-entry
quarters and across all incumbents in route i
IncPrePlaneSizei An average number of aircraft seats over 8 pre-entry quarters
and across all incumbents in route i.
Po
st-e
ntr
y M
ark
etin
g
Va
ria
ble
s
IncPostPriceCutij Market-share weighted average price-cut across all
incumbents in quarter j post-entry divided by IncPrePricei.
IncPostFreqij Market-share weighted average flight frequency across all
incumbents in quarter j post-entry divided by IncPreFreqi.
IncPostPeakFreqij The market-share weighted average percentage of flights
during peak hours across all incumbents in quarter j post-entry.
IncPostOTPij The market-share weighted average percentage of on-time
flights across all incumbents in quarter j post-entry.
Co
ntr
ol
Va
ria
ble
s
Hubi Equals 1 if one of the endpoint airports of route i is an
incumbent’s hub, 0 otherwise.
Distancei Distance between two endpoint airports of route i in 100 miles.
(MMCij)
A number of routes within the challenger’s network where the
challenger faces the same incumbents in route i, divided by the
challenger’s number of routes.
ChllgSizeij A total number of passengers in quarter j traveling with the
challenger that entered route i over its entire network.
Demandi The geometric mean of the population of the endpoint cities in
route i.
ChllgNetworkij A number of routes in quarter j that originate from the two
endpoints of route i divided by the challenger’s network size.
IncNetworkij A number of routes in quarter j that originate from the two
endpoints of route i divided by the incumbents’ network size.
NIncij Number of incumbents in route i in quarter j.
2ndEntryi Equals 1 if a 2nd challenger entered route i while the first
challenger is still in the route, 0 otherwise.
FuelPricej Price of fuel in quarter j (dollars per gallon).
44
TABLE 1.3: Split-population Model Results
Coef. S.E.
Time-To-
Exit
(Hazard
Regression)
Incumbents post-entry
marketing
Price-cut -1.43601 0.18469
Flight frequency 0.32821 0.07477
Peak flight frequency 0.17922 0.25985
Flight OTP -1.51105 0.33487
Flexible time function ×
Incumbents post-entry
marketing
t × Price-cut -0.00214 0.01545
t × Flight frequency 0.03080 0.01174
t × Peak flight frequency -0.08110 0.02861
t × Flight OTP -0.07941 0.04425
t2 × Price-cut -0.00023 0.00015
t2 × Flight frequency -0.00043 0.00015
t2 × Peak flight frequency
frequencyfrequency 0.00105 0.00033
t2 × Flight OTP 0.00099 0.00054
Ln(t) × Price-cut 0.61842 0.14333
Ln(t) × Flight frequency -0.24345 0.06873
Ln(t) × Peak flight frequency
frequency 0.20185 0.22261
Ln(t) × Flight OTP 1.23556 0.30833
Controls
Low-cost
challenger
Price -0.00137 0.00036
Size 0.00009 0.00001
Market-level
characteristics
2nd Entry 0.22832 0.06477
Hub -0.00308 0.02649
Distance 0.00965 0.00286
Multi Market Competition 0.36974 0.11589
Fuel Price 0.10357 0.01757
Route Demand -0.20187 0.02997
Number of Incumbents 0.10274 0.02882
Network-level
characteristics
Challenger Route Importance -1.56558 0.19847
Incumbent Route Importance 0.00275 0.00049
Exit
Likelihood
(Logit
Regression)
Incumbents pre-entry
marketing
Price -0.00936 0.00367
Flight Frequency -0.00327 0.00139
Peak flight frequency 1.67274 1.55800
Flight OTP -2.43487 1.81448
Plane Size 9.20489 0.0118
Low-cost
challenger Price -0.00860 0.00345
N=13057; *p < .10, **p < .05, ***p < .01; Shape parameter = 1.624*** (S.E. = .0527), AIC =
3,015.9, BIC = 3,494.4. Notes: Intercept estimates are removed from the table for the sake of space
(Intercept Prob. of exit = -5.36** (2.218), Intercept Time-to exit = -1.78*** (.122)). Three incumbent
Fixed Effects (US, DL, NW) are significant at 5%, all challenger dummies (WN, B6, FL, F9) and
are significant. All year-dummies are also significant at 1%. UA: United Airline, AA: American
Airlines, US: US Airways. DL: Delta Airlines, NW: North West Airlines, CO: Continental Airlines,
WN: Southwest Airlines , B6: JetBlue Airways, FL: AirTran Airways, F9: Frontier Airlines.
45
TABLE 1.4: Robustness Checks
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Exit
Likelihood
Pre-entry Price -0.00876 -0.00946 ** -0.00814 * -0.01332 ** -0.00167 -0.01487
Pre-entry Frequency -0.00119 -0.00197 -0.00163 * -0.00732 ** -0.00138 0.00867 *
Pre-entry Peak frequency 1.25812 1.85325 0.47514 1.22358 1.82955 -8.28377 *
Pre-entry flight OTP -10.38112 -8.70457 *** -7.98547 *** -6.83955 *** -10.65558 *** 5.66677
Pre-entry Plane Size -0.02762 * -0.04421 *** -0.03680 *** -0.02715 *** -0.02707 * 0.00653
Challenger Price -0.00916 -0.00455 -0.00299 -0.00454 -0.00817 0.04867 **
Time-To-
Exit
Post-entry price-cut -1.42000 *** -1.67152 *** -1.78225 *** -1.41558 *** -1.43000 *** -2.05875 *** Flight frequency 0.44424 *** 0.30221 *** 0.28158 *** 0.28258 0.17921 0.23136 ***
Peak flight frequency 0.36352 0.23774 0.08891 0.05315 0.46611 * -0.47258
Flight OTP -1.11948 *** -2.42433 *** -4.02122 *** -1.42025 *** -1.33414 *** -0.76858 *
t × Post-entry price-cut -0.00561 -0.01416 -0.01991 -0.05151 0.00132 -0.25722 ***
t × Flight frequency 0.04065 *** 0.01685 0.01297 0.03885 * 0.00799 0.07235
t × Peak flight frequency -0.03080 -0.08738 *** -0.11425 *** -0.10225 *** -0.02944 -0.26647 *
t × Flight OTP 0.00924 -0.20165 ** -0.43774 *** 0.04425 0.07514 * -0.11885
t2 × Post-entry price-cut -0.00015 -0.00001 -0.00004 0.00048 0.00039 ** -0.00491
t2 × Flight frequency -0.00050 *** -0.00023 -0.00019 -0.00072 *** -0.00008 -0.00258
t2 × Peak flight frequency
timesq
0.00029 0.00118 *** 0.00154 *** 0.00143 *** 0.00017 0.00455
t2 × Flight OTP -0.00014 0.00288 ** 0.00611 *** -0.00104 * 0.00117 ** -0.00071
Ln(t) × Post-entry price-cut 0.60847 *** 0.76458 *** 0.83814 *** 0.84652 *** 0.69958 *** 1.83025 ***
Ln(t) × Flight frequency -0.33754 *** -0.18621 * -0.16352 * -0.26325 * -0.10775 -0.30158 **
Ln(t) × Peak flight frequency -0.05644 0.18811 0.35135 0.35425 -0.11912 1.00355 **
Ln(t) × Flight OTP 0.72612 ** 2.13832 *** 3.81745 *** 0.75845 ** 1.06125 *** 0.22487
Challenger Price -0.00121 *** -0.00134 *** -0.00143 *** -0.00035 -0.00150 *** -0.00050 **
2ndEntry 0.12124 *** 0.24825 *** 0.25887 *** 0.28912 *** 0.14925 *** 0.11758 ***
Hub 0.01348 -0.00983 -0.02516 -0.02422 -0.02563 -0.02137
Distance 0.00046 0.00781 ** 0.00787 ** 0.00102 0.00027 0.00103
Multi-Market Competition 0.06294 0.34815 *** 0.27345 ** 0.14802 0.39525 *** 0.14114
FuelPrice 0.08886 *** 0.10624 *** 0.10485 ** 0.20522 *** 0.09454 *** 0.06634 ***
Challenger Size 0.00009 *** 0.00011 *** 0.00011 *** 0.00004 ** 0.00011 *** 0.00005 ***
Route Demand -0.14145 *** -0.19225 *** -0.19421 *** -0.07122 ** -0.15225 *** -0.08235 ***
Challenger Route Importance -1.34168 *** -1.61357 *** -1.60457 *** -0.83912 *** -1.50454 *** -0.95885 ***
Incumbent Route Importance 0.00293 *** 0.00249 *** 0.00245 *** 0.00061 0.00255 *** 0.00151 ***
Number of Incumbent 0.06677 ** 0.11225 *** 0.10015 *** 0.01221 0.11511 *** 0.05644 ***
Observations 16209 13057
13057 7689 9754 7922
46
ESSAY 2: BLUFF OR REAL? HELPING INCUMBENTS RECOGNIZE
HOW A REALLY THREATENING FIRM LOOKS LIKE17
“… All warfare is the way of deception. Offer the enemy a bait to lure him; create havoc in the east and strike in the west …” (Tzu 1963)
Informed by probability and psychology, in the opening poker scene of the 1965 movie
“The Cincinnati Kid”, Steve McQueen’s character is able to read the bluff in Buster’s face
(his opponent) and – instead of dropping the game and against everybody’s disbelief –
make the call and collect ‘all the cash in the pot.’ Bluff which is defined as strategic
deception, is a tactic that poker players, military generals, or politicians use – and so do
managers – to mislead opponents about their true plans (Hendricks and McAfee 2006). In
business as in poker, “bluff is a common strategic move” used to influence competitors to
take, or not take, a specific course of action that leaves them worse off (Porter 1980; Prabhu
and Stewart 2001).18
Besides well-known ‘verbal bluffing’ using pre-announcements,19 firms can also
deceive competitors by means of their observable moves (Prasad Mishra and Bhabra 2001).
For instance, firms often use maneuvers to camouflage their true intentions and plans when
trying to enter a new market (McGrath, Chen and MacMillan 1998; Hendricks and McAfee
17 Aghaie, Sina., Carlos Lourenço, Charles Noble and Rafael Arreola. To be submitted to Marketing Science 18 See https://insight.kellogg.northwestern.edu/article/to-bluff-or-not-to-bluff. Companies may also use bluff
on other stakeholders such as customers (Porter 1980; Prabhu and Stewart 2001). 19 In the software industry, Microsoft Inc. frequently announced new products that never appeared on the
market (Prabhu and Stewart 2001) and Intel and Motorola have been accused of announcing “vaporware”
product (see https://www.forbes.com/2007/01/04/vaporware-ces-gadgets-tech-media-
cx_rr_0105vaporware_slide.html#29f9ffdb3154
47
1998). Using this strategy, a potential entrant (PE hereafter) may threaten multiple markets
to increase an incumbent firm’s uncertainty about which market is, and which one is not, a
real target market for the PE. In this situation, an incumbent firm that usually reacts to these
entry threats to deter future entry (Ellison and Ellison 2011; Homburg et al. 2013) is forced
to allocate its limited resources to the multiple markets that are, possibly, under threat.
Consequently, it will have fewer resources to react when a market entry does occur, and
the absence of resources will reduce the incumbent’s ability to retaliate at a time the PE is
most vulnerable in a new market. In other words, by better predicting a real threat vs. a
bluff, an incumbent could mitigate over- and under-reaction to market-entry threats.
Usually, incumbent firms face several market entry threats simultaneously, but
because resources are limited, they cannot respond to every one of those threats: they would
like to ignore irrelevant threats and react to only important ones. Thus, recognizing the type
of threat a firm is facing is one of the most crucial decisions in marketing and managers
should identify those threats that are real and that deserve an appropriate response (Klemz
and Gruca, 2003). In the multi-billion-dollar airline industry, for example, 15% to 20% of
threats turn into an actual entry (Gayle and Wu, 2013; Parise 2018), a sizeable proportion
that speaks well to the managerial implications of identifying the markets under a real
threat of entry. In line with that promise – that threat type identification may improve an
incumbent’s performance by enhancing its resource allocation efficacy – prior research
(Eliashberg et al. 1996) recommended to “study how effectively managers distinguish
between bluffs and truthful signals” (p. 31). Surprisingly, however, and despite a long-
lasting call to close this gap (Chen and Miller 2012; Eliashberg et al. 1996) and the
prevalence of competitive bluffing among market players (Guidice et al. 2009), to the best
48
of our knowledge, there is no empirical study that we are aware of on the drivers of real
threat vs. bluff.
In this paper, we take a step at addressing this gap in the marketing strategy
literature by empirically investigating the characteristics of a real vs. bluff threat. More
specifically, we estimate the probability of a real vs. a bluff as a function of (i) market
characteristics, (ii) the characteristics of both the potential entrant and the incumbents, and
(iii) the market network structure of both firms. According to the awareness-motivation-
capability framework (Chen et al. 2007), we argue that threats are more likely to be real
when the potential entrant has the motivation to enter a market as well as the capability of
doing so. This is challenging, however, because real threats vs. bluffs are not entirely
observed — that is, there are missing values in the dependent variable. To overcome this
challenge, we propose to employ a multiple imputation chained equation (MICE) method
that makes use of two fundamental pieces of information: observed market entries and
incumbents’ marketing reactions (i.e., price-cuts). This method can simultaneously impute
the missing values for the threat type and estimate the effects of the variables of interest on
the likelihood of the threat being a real one.
Market entry threats posed by low-cost firms are a recent reality in many industries,
and a well-known phenomenon in the airline industry for quite some years now (Ethiraj
and Zhou 2019; Goolsbee and Syverson 2008), thus offering researchers an ample time
window to study the impact of low-cost market entries and threats. We test our hypotheses
empirically on an extensive, multi-market dataset from the US airline industry. This
industry is particularly well suited for our purposes because threats of entry are frequent
and easily observed, and the identification of potential entrants and existing incumbents is
49
well established (Claussen, Essling and Peukert 2018; Ethiraj and Zhou 2019; Goolsbee
and Syverson 2008; Prince and Simon 2014).
For managers of incumbent firms, our findings may help to implement effective
preemptive strategies when facing entry threats by potential entrants. In particular, our
study adds to our understanding of resource allocation – and its management in a
competitive environment. From an academic point of view, since not all threats are actual
threats to incumbents in a market (Gayle and Wu 2013), we contribute to prior literature
on market entry threat (Aydemir 2012; Ethiraj and Zhou 2019; Goetz and Shapiro 2012;
Goolsbee and Syverson 2008; Prince and Simon 2014) by exploring the factors that can
explain the seriousness level of entry threats posed by potential entrants. Moreover, we
consider PE’s motivation and capability as drivers of its market entry decision, which
typically have not been studied jointly in the previous literature.
The paper is organized as follows. First, we discuss the related theoretical
background. Second, we develop a conceptual framework linking a threat type to the
resources and capabilities of a potential entrant and of incumbents and derive predictions
based on this framework. Third, we discuss our empirical modeling and estimation
strategies and describe our airline industry data and the operationalization of the different
variables used. Finally, we present the results and discuss their implications as well as the
limitations of our study and future research opportunities.
THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT
The entry of new firms into an existing market increases competition, hurts incumbents’
market shares, and erodes their profits (Geroski, 1995). Given the potentially disruptive
effects of a new entrant on incumbents’ performance (Parise 2018), incumbents respond
50
quickly and forcefully to threats (Goetz and Shapiro 2012; Goolsbee and Syverson 2008;
Parise 2018; Seamans 2013) to deter entry (Cookson 2017; Dafny 2005; Ellison and
Ellison, 2011; Ethiraj and Zhou 2019). However, not all threats posed by these potential
entrants are the same, and only serious threats should make incumbents respond and justify
the use of price- and/or capacity-driven competitive resources (Ethiraj and Zhou 2019;
Gayle and Wu 2013; Wang et al. 2016). This is easier said than done, since incumbents
need to first and foremost identify the type of threat they are facing, i.e. whether it’s a
serious or credible one or simply a bluff.
Incumbents are not always successful drawing a distinction between a serious or
real threat and a bluff (Karaer and Erhun 2015). Given that competitive responses (e.g.,
price reduction, capacity expansion) are often costly to incumbents (Heil and Robertson,
1991), resource misallocation in response to threats has a detrimental impact on the
incumbents’ performance and ability to defend their markets. Surprisingly, prior literature
left unexplored how an incumbent can more accurately draw a distinction between serious
threats and bluffs. Relying on the awareness-motivation and capability (AMC) framework
(Chen 1996; Chen et al. 2007), we explore the firm and market level correlates of a real vs.
a bluff threat.
Awareness, Motivation, and Capability Perspective
According to the AMC framework (Chen 1996; Chen et al. 2007), three behavioral drivers
influence a PE’s decision to enter a market: awareness, motivation, and capability. Thus, a
PE needs to not only be aware of the markets it wishes to enter, but also be motivated to
and capable of doing so. That is, awareness may make a PE threaten a market, but it is
essentially motivation and capability that turn a threat into an actual entry – and lack of
51
motivation and/or capability turn a threat into a mere bluff. In the context of the US airline
industry, we assume all market players are aware of all existing markets, which are the
officially authorized routes linking national airports (this is the common definition of a
market in the industry and in past literature; see e.g., Ethiraj and Zhou 2019; Prince and
Simon 2014)
Since a rival’s move is best seen through the rival’s eye (Tsai et al. 2011), an
incumbent observing a threatening move from a PE firm should concentrate on figuring
out whether the PE really is motivated and/or is capable of actually entering the market.
Accordingly, we discuss in detail next the underpinnings of a PE’s motivation and ability
to attack and enter a new market.
Motivation to Attack and Enter a New Market
First and foremost, a PE’s perception of market attractiveness (e.g., market demand,
growth, competitive intensity, etc.) influences its motivation to enter a market and the PE
is more likely to attack markets that are highly attractive (Dixit and Chintagunta 2007). But
although market attractiveness is the crucial motivational factor that influences entry
decision, not all attractive markets are worth attacking. A PE also evaluates the risk
associated with the market entry (Clark and Montgomery 1998), regardless of how
attractive the market is. Expected attractiveness is not only a function of market demand
and a market growth rate but is also a function of how easily the PE can capture the
expected demand. Specifically, markets where incumbents have high resource
redeployment capabilities will be riskier to attack because capable incumbents can easily
and swiftly redeploy their resources to defend the market, thereby reducing the chance of
success for newcomers. Thus, we argue that the incumbents’ capability and available
52
resources would definitely influence PE’s motivation to entry. Moreover, a market’s
importance to the incumbent is an important signal of a market’s attractiveness to the PE:
as the importance of the market increases to the incumbents, they are more likely to defend
the market at the time of rival’s entry and make the entry a riskier move for the PE. Thus,
route importance may also influence PE’s motivation. Given all the above, we hypothesize
that:
H1a: The higher the market attractiveness, the more likely that the posed threat by
the PE is a serious one.
H1b: The higher the incumbents’ available resources (capabilities) in the market, the
less likely that the posed threat by the PE is a serious one.
Firms often compete against each other in many markets simultaneously. This multimarket
competition (MMC) influences the competitive behavior of firms (Baum and Korn 1996;
Gimeno 1999; Jayachandran, Gimeno, and Varadarajan 1999), in the sense that the higher
the number of markets where the PE and the incumbents compete, the softer the intensity
of their competitive activities (Baum and Korn 1999; Prince and Simon 2009). An
incumbent is thus less likely to defend its markets at the time of entry (Ethiraj and Zhou
2019) if it competes with the PE in several markets already. Therefore, the PE’s hesitation
to attack the market would be lower. Accordingly, we hypothesize the following
moderating effect of MMC:
H2: The MMC between PE and incumbents weakens the negative correlation
between incumbents’ capabilities and the likelihood that the posed threat by the PE
is a serious one.
Ability to Attack and Enter a New Market
While having motivation is a necessary condition for market entry, it is not a sufficient
one. A motivated PE unable to attack the market does not pose a serious threat and is less
53
likely to enter a new market no matter how attractive the market is. A PE’s ability to attack
highly depends on its resources, both core and complementary (Helfat and Lieberman
2002). Core resources include financial or physical assets, technological knowledge and
knowledge of customer needs and complementary resources include customer service,
distribution and logistics, and marketing and sales. We predict that the higher the PE’s
ability the more its threatening moves should be regarded as signals of real threats, rather
than the bluff by market incumbents. More formally, we propose that:
H3: The higher the potential entrant’s (PE’s) resources and capabilities, the more
likely that its posed threats are serious ones.
The conceptual framework, depicted in Figure 2.1, suggests that not only incumbents’ but
also potential entrants’ resources and capabilities, as well as marketplace characteristics,
determine the nature of the posed threats by the PE. Implicitly, we assume incumbents can
reduce the uncertainty regarding whether a competitive move is either a real threat or a
bluff by ‘reading potential entrants’ faces,’ i.e., by taking into consideration PEs actions
and characteristics.20 Next, we test the hypotheses discussed above on a large scale,
longitudinal dataset on threatening moves by low-cost PEs in the airline industry.
DATA, INDUSTRY CONTEXT, AND THREAT CLASSIFICATION
The Airline industry is particularly well suited for our purposes because each one of the
thousands of routes between any two airports is considered a unique market (Claussen,
Essling and Peukert 2018; Dixit and Chintagunta 2007; Ethiraj and Zhou 2019; Prince and
Simon 2014), where entries and entry threats are frequent and easily observed, and the
20 From an econometrician’s point of view our uncertainty regarding the identification of a real threat vs. a
bluff is lower than that of incumbent firms at the time they face their PEs’ actual moves. In fact, by observing
past competitive moves of incumbents and their PEs and making a few assumptions, we can, not without
limitations, generate the threat vs. bluff data, as explained in the text.
54
identification of the potential entrants and existing incumbents is well established (Ethiraj
and Zhou 2019).
In this research, we explore threats posed by low-cost carriers (LCC), which are
frequent in this industry – and virtually all airlines will at some point face low-cost
competitors (Ethiraj and Zhou 2019; Gerardi and Shapiro 2009; Parise 2018).21 Focusing
on threats posed by LCCs – as opposed to major carriers – is important because, over the
past three decades, low-cost carriers have significantly increased their domestic market
share and have entered major airlines’ markets (Ethiraj and Zhou 2019). Furthermore, the
entry of a low-cost carrier has a much larger impact on incumbents’ profit margins than
the entry of a major carrier (Parise 2018). Also, unlike major airlines that have alliances
and code sharing with each other, low-cost airlines usually avoid such collaborations,
further making their moves be taken as threatening from a competitive point of view (Goetz
and Shapiro, 2012). Finally, major airlines take not only route- but also the entire network-
profitability into account when deciding to get into or stay out of a particular route, and the
focus on LCCs avoids this type of confounds.
Our data cover market-level information, carriers’ characteristics, and firms’
activities, from the first quarter of 1997 through the fourth quarter of 2015. Five low-cost
carriers, AirTran, Southwest, JetBlue, Frontier, and Spirit have remained significant
players in the U.S. airline industry throughout that period.
Threat definition. Before discussing our dependent variable, i.e., whether a threat
is a serious one or a bluff, we first define a threat per se. To determine whether a market
21 When analyzing firms’ decisions to enter a market, sunk costs may be a confound factor. Since it is typically
unavailable to researchers, it would be difficult to empirically control for it (Dixit 1989; Elfenbein and Knott
2015; O’Brien and Folta 2009). In the airline industry, however, sunk costs are negligible (see also Aghaie,
Lourenço, and Noble 2019; Cabral and Ross 2008).
55
or route i is under a threat of entry by a potential entrant, we rely on Goolsbee and Syverson
(2008) definition of “entry threat.” In a given route, if a low-cost carrier is operating flights
out of both endpoint airports of that particular route but is not actually operating a nonstop
flight on that route, the route is under the threat of entry by that LCC. As an example,
consider that an incumbent is serving the route between Miami (MIA) and Washington
(IAD) at T=0. Imagine that an LCC starts flying out of Miami (MIA) to Denver (Den) at
T=1. Although the LCC got close to the MIA-IAD route, it still does not threaten that
market. Imagine further that the LCC enters the Washington Dulles International Airport
(IAD) and starts operating between IAD and other airports (e.g., Atlanta (ATL)), just not
the IAD-MIA route. According to Goolsbee and Syverson (2008), once an LCC operates
out of both airports of one route – in this example, MIA first and then IAD – the probability
that the LCC soon starts serving the route itself increases dramatically. In other words,
once the LCC establishes its presence at the second airport of the market, it poses an entry
threat to the incumbents on that market, i.e., the market is under threat by the LCC (Please
see Figures 2.1 a-c).
Goolsbee and Syverson (2008) argue that when a low-cost airline has gates,
counters, ground crew and maintenance facilities established at both airports of a particular
route, it would be easier to begin nonstop service on the route between two airports. We
should highlight that the threat-of-entry proxy is only appropriate for LCCs due to the way
in which these airlines are willing to fly routes between two non-hub airports. Parise (2018)
showed that once the low-cost potential entrant establishes its presence at the second
endpoint airport of route i, the probability of the actual entry in route i increases by 36%.
56
For traditional hub-and-spoke major airlines22, however, the mere presence of operations
in two airports is not a meaningful predictor of future nonstop service between the
endpoints, since hub considerations are far more critical for such carriers (Aydemir, 2012;
Goetz and Shapiro, 2012).
Threat type classification. Our main goal is to understand how market-specific
factors, together with incumbents’ and a PE’s characteristics, affect the likelihood that
incumbent firms are faced with a serious threat vs. a bluff. To classify threats into one of
the two types of the threat we use two fundamental pieces of information: incumbents’
observed marketing reactions to the threat (if any) and the PE’s entry (if any). Imagine a
potential entrant’s action brings it closer to an incumbent’s market. Imagine further that an
incumbent does not react to this action. In that case, if, despite the incumbent’s apathetic
approach to the PE’s threat, the PE does not enter the market anytime soon, the move was
likely to be a bluff (and maybe it has been a bluff since then) – and the incumbent was right
doing nothing. But if the PE soon enters the market, the move was a real threat back then
– and maybe the incumbent should have done something about it. Now imagine an
incumbent does react. If the PE soon enters the market, the move was a real threat – and
one the incumbent could not avoid despite trying. The real challenge is how to classify
cases where the incumbent reacts to the threat of entry and we do not observe a subsequent
entry anytime soon. It is hard to tell what type of move that was back then – it could be
that either the incumbent’s reaction deterred the PE’s entry, or the opponent was bluffing
22 Hub and spoke’ systems connect origins and destinations through hubs. For example, passengers from one
city with different destinations are carried together on a flight to a hub (this flight or route is called ‘spoke’).
Then, they are combined with passengers arriving from other cities into a hub and finally this passenger pool
will be regrouped onto separate flights (spokes) to different destinations. High traffic volume at hubs allows
firms to take advantage of the economies of scale in origin-hub and hub-destination flights (see Pirkul and
Schilling 1998).
57
all along, i.e. the PE was, ‘in reality’, not planning to enter the market. In sum, we can
derive the ‘data’ for our dependent variable ex-post in all possible cases, but one (Table
2.1 a and b). We propose to handle this one case as a missing data problem.
EMPIRICAL ANALYSIS
Having a binary dependent variable would suggest a discrete choice model such as probit
or logit. However, to be able to include several time- and firm-fixed effects into our model,
prior studies (Claussen, Essling, and Peukert 2018; Hellevik, 2009) suggest applying a
linear probability model (LPM). Moreover, compared to the non-linear models, LPM
allows researchers to easily provide a more meaningful interpretation of the main effects
and the interaction terms. We ran both linear and logistic models. Since the sign and
significance of the coefficients of interest, and therefore our overall results, are virtually
the same, we only discuss the LPM results (all results are available upon request).
Model
We model the likelihood π that a PE’s threat is a serious one (hence 1- π is the likelihood
that the threatening move is a bluff) as a function of the potential entrant’s motivation and
capability of market entry, incumbents’ characteristics, and market-specific factors.
Specifically, π(real) in route i is a function of market characteristics (e.g., demand, growth
rate, competitive intensity) and the PE’s and incumbents’ resources and capabilities. Thus,
we develop a linear regression model as follows:
π(reali)= β0 + β1IncR&Ci + β2MarketGrowthi + β3MarketDemandi + β4Pricei +β5Delayi +
β6Importancei + β7PE-R&Ci + β8PE-Size + β9FuelPrice + β10Distancei + β11DistanceSQi +
58
β12MMCi + β13NInci + β14Leisurei + β15Hubi + β16LoadFactori + β17MMCi×PE-R&Ci + β18-
21PEi + β22-37Yearj + β38-47INCi (2.1)
All measurements are averages over eight pre-threat quarters. The right-hand side
independent variables are operationalized as described below.
PE’s Motivation to Enter.
Incumbent’s Resources & Capabilities (R&Cs). Airport level investments can be a good
proxy for route-level investments (Prince and Simon 2014). For instance, carriers can hire
more employees to speed up several processes at the airport such as loading and unloading
baggage, check-in, etc. Moreover, airlines can have an additional airplane at an airport or
have a ready supply of mechanics available to avoid any issue resulting from unexpected
mechanical failures. More importantly, airlines can acquire more gates and increase the
number of counters at airports. Moreover, since available R&Cs at the airport (e.g., number
of aircraft, counters, gates, employees) would be easily redeployed to any route that
originates from the airport, airport level resources would be highly correlated with the route
level R&Cs. Thus, we use the average of the incumbent’s R&Cs at the endpoint cities as a
proxy for the incumbent’s available R&Cs at each route. For instance, for the route O-D,
the incumbent’s R&C would be (OR&C + DR&C)/2. For routes with more than one
incumbent, we use market share weighted averages to calculate route level R&Cs. The use
of weights based on market shares ensures that the relative competitive strength (leader vs.
followers) of incumbents in a market, and their impact on potential entrant’s decision,
remains23.
23 The use of market-share weighted averages assumes the potential entrant looks at the resources of a
‘representative incumbent’ while still preserving market-share differences. In other words, the resources
59
Airport level R&Cs measurement. Since all of these investments plus airline’s
technical and managerial capabilities would be reflected in the airline’s flight schedules,
frequency, and on-time performance, we use flight frequency and schedule as a proxy for
the incumbent’s resource and capabilities at the airport.24 Accordingly, we develop six
measures to capture different aspects of airport level R&Cs. The first of the incumbents’
R&Cs measure at airport i, is the maximum number of non-stop flights per quarter
(MaxQtri) that land or depart the airport during the pre-threat stage. As a second measure,
we generate MaxDayi which is the maximum number of non-stop flights (inbound or
outbound) per day across the pre-threat period and, as a third measure, we generate
MaxHouri which is the maximum number of non-stop flights per hour across the pre-threat
period. Carriers also compete with each other to offer consumers more convenient access
to their service by increasing the frequency of flight departures in peak times. Prior
literature defines a peak period as a 7-10am or 3-7pm on weekdays (see Oliveira and Huse
2009; Sengupta and Wiggins 2014). Peak time-frequency affects passengers’ choice of the
airline because travelers are both price- and time-sensitive (Shaw 2007). Accordingly, the
other three measures for airlines’ airport R&Cs, are MaxPeakDayi, MaxPeakQtri, and
MaxPeakHouri which are the maximum number of non-stop peak time flights (inbound or
outbound) per day, per quarter, and per hour, respectively. Since these six measures are
highly correlated and capture different aspects of firms’ R&Cs, we compute incumbents’
R&Cs at each airport as the principal component score of the above six indices. Factor
possess by say an undisputed market leader will show more strongly than those with negligible market shares.
In such cases, a potential entrant is likely to pay more attention to ‘who possesses what’. 24 Incumbent’s resource and capability would also influence its On-Time performance. However, since a big
portion of delays might be due to the other airport-level factors that are out of the airline’s control, we decided
to consider incumbent’s on-time performance as a control variable.
60
analysis confirms that a single factor accounts for 99% of the six scores’ combined
variance. The composite variable is standardized to have a mean of 0 and a standard
deviation of 1.
Market importance to the incumbent. In network industries such as the airline
industry, in which firms operate and interact with each other in several interconnected
markets, the importance of a market for a given firm is related to the firm’s market network
structure. This is because what happens in one market is not entirely independent from all
other markets, and thus the perceived importance of each market should be evaluated not
only by its stand-alone appeal but also by its connection with other markets. In the airline
industry, the different (geographical) markets are naturally connected by the very nature of
routes linking any two airports, and some routes are more central (important) than others.
We measure incumbents’ route importance or route centrality within their networks,
IncNetwork using a measure developed by Dunn (2008): for each route, the network
importance measure is determined by the number of non-stop markets that originate from
the two endpoints (excluding the non-stop route to the city being considered) divided by
its network size. For instance, if, in a route between city “O” and city “D”, an LCC has five
non-stop routes out of “O” and six non-stop routes out of “D”, and it serves 100 routes
within its network, then the network centrality (importance) of route O-D is [(5+6)-2] / 100
= .09.
Market demand and growth rate: we measure route demand as a geometric mean
of the population in the endpoint cities in the pre-threat period (Dixit and Chintagunta’s
2007). This variable is important because PEs usually focus on serving routes with high
potential demand. The higher the mean number of passengers, the greater is the expected
61
attractiveness of the market. And the market growth is the average quarterly growth rate
across the pre-threat period.
Pre-threat environment: we focus on two fundamental aspects of the competitive
environment before threats unfold (of any competitive environment for that matter): market
price characteristics and service quality levels. To capture price characteristics (Pricei), we
first calculated the incumbents’ weighted average of prices, price variances and median
prices over eight pre-threat quarters, where incumbents’ market-shares serve as weights.
Then, we performed principal component analysis on these three measures and generated
a univariate score to measure price characteristic. Since higher scores indicate that the
incumbents charge high prices with high variances, the PE’s low-cost proposition would
be a particularly compelling one among price-sensitive consumers that higher priced
mainstream carriers are not serving effectively. As a result, the market with higher price
score is a good target for the PE.
One of the main indicators of service quality in the airline industry is the percentage
of flights that arrive on-time (Grewal, Chandrashekaran, and Citrin 2010), which is
available at route-level (see Prince and Simon 2014). Prior literature proposed three
different definitions for the delayed flight; (1) if the flight arrives at least 1 minute late, (2)
if it arrives at least 15 minutes late, and (3) if it arrives at least 30 minutes late. Therefore,
to develop a measure for OTPi, we first calculated the market-share weighted average of
the percentage of flights that are late using these three thresholds (Prince and Simon 2014).
Then again, we performed principal component analysis on these three measures and
generated a univariate score to measure the service quality. The expected sign of this
variable is not clear, on the one hand, higher delay may encourage the PE to attack the
62
markets with the inferior service quality. On the other hand, since high market level delay
might prevent the PE from keeping the turnaround time as low as possible, the PE may not
be interested in attacking those targets.
Multi-market competition (MMC). Airlines often compete against each other in
many markets simultaneously, which influences competitive behavior (Baum and Korn
1999; Gimeno 1999), thus we define a multimarket variable (MMC), as follows. For PE ,
in route i, we count all common routes with incumbents over all routes in quarter j that
threat starts and then divide the PE’s total contact by (n - 1), where n is the number of
incumbents in route i. Finally, we standardized the average count by the number of markets
served by the potential entrants in quarter j (for a review of MMC operationalizations, see
Baum and Korn 1996).
PE’s Ability to Enter
PE’s Resources & Capabilities (R&Cs). Similar to what we did for incumbents, we
measured a PE’s MaxQtri, MaxDayi, MaxHouri MaxPeakDayi, MaxPeakQtri and
MaxPeakHouri at each airport and then generated a principal component score of these six
indices as a measure of that PE’s R&C’s at each endpoint city. Finally, the PE’s R&C
would be (OR&C + DR&C)/2.
PE’s size. Well-established PEs with a large national network may have higher
resources and capabilities and are probably more successful than smaller PEs in pursuing
their growth strategies. For example, bigger PEs are stronger financially (Claussen,
Essling, and Peukert, 2018), have more experiences and infrastructures and therefore may
have more staying power in, say, price wars (Dixit and Chintagunta 2007). Furthermore,
larger PEs may simply enjoy high brand recognition, higher operational experience, or
63
logistical advantages of larger networks (Ito and Lee 2003). Thus, larger PEs would be
more likely to pose a serious threat than smaller ones would. In our model, we use the
natural log of the total number of passengers that are carried by the PE as a measure of the
PE’s size (Dixit and Chintagunta 2007).
Fuel cost. Low-cost airlines are by nature more vulnerable to fluctuations in
production costs than legacy airlines. Since fuel cost is one of the most important expenses
for airlines in general – it accounts for almost 30% of operating costs – its fluctuations are
likely to seriously limit LCCs capabilities to growth. For instance, in 2008 a 7% increase
in fuel price has been estimated to decrease an airline’s net profit by almost 2%. During
the time of our study, the fuel price had a large variation (±30% to ±80% year-over-year)25
and because the low-cost airline is more vulnerable to these variations, we believe that the
higher fuel prices can also reduce PE’s motivation to enter a market.
Control variables
A wide array of factors may also influence the type of posed threat that should be controlled
for. We use control variables related to firms as well as market level controls.
Competitive intensity and Route distance. NInci is the total number of incumbents
in route i, and Distancei the distance between two endpoint airports for each route. Short
haul routes are more attractive than long ones for LCCs because their cost structure requires
a quick turnover (Berry and Jia 2010). This means LCCs have a higher incentive to target
25 See https://www.iata.org/pressroom/facts_figures/fact_sheets/Documents/fact-sheet-fuel.pdf for an
estimate of fuel price’s weight on operating costs. The estimated impact of fuel price on profits can be found
at https://www.forbes.com/sites/greatspeculations/2015/07/29/american-airlines-profit-surges-on-fuel-cost-
savings-unit-revenue-to-remain-weak-until-mid-2016/#14cf6fec6d4d. In our data, the average fuel price was
0.35$/gallon in 1998, 1.4$/gallon in 2004, 3.89$/gallon in 2008, 1.3$/gallon in 2009 and 3.1$/gallon in 2011.
64
short-haul routes. Therefore, in each route, we control for the distance between origin and
destination airports, in miles.
Load factor. The passenger load factor measures the capacity utilization of airlines
and somehow indicates how efficiently the airline is performing. We expect that potential
entrants are less likely to target routes in which incumbents have high load factors. We
operationalize a load factor as the percentage of available seating capacity that is filled with
passengers.
Market type. Since leisure travel demand is more price sensitive, LCCs target
markets with a high percentage of leisure passengers (Bendinelli et al. 2016). We identified
leisure routes using Gerardi and Shapiro’s (2009) list of leisure destinations in the US. If
one of the route endpoints is among these leisure cities, we coded the route as a ‘leisure
route.’ Therefore, the leisure variable equals one for the leisure route and zero otherwise.
Firm, year and market fixed effects. Finally, we include a set of yearly dummies
Yearj to capture unobserved time-varying macroeconomic factors such as shifts in demand
and costs of production, and other unobserved time factors (Greenfield 2014; Mayer and
Sinai 2003), and a set of potential entrants, PEi and incumbent dummies, INCi, to capture
potential unobserved incumbent- and potential entrant-specific factors. Any inherent
differences between PEs that might influence the type of threat they impose are therefore
captured by these fixed effects. See Table 2.2 for a summary of descriptive statistics of all
variables and their correlation matrix.
Estimation
As mentioned before, we handle the cases where the incumbent reacts to the threat of entry
and we do not observe a subsequent PE entry as cases of missing data because it could be
65
that either the incumbent’s reaction deterred the PE’s entry, or the opponent was bluffing
all along.
We used three different strategies to handle the missing observations and estimated
different models. Initially, we assume that observations are missing completely at random
(MCAR). Second, we mean imputed the missing IVs observations and only dropped the
missing DV observations. Finally, we estimated a multiple imputation chain equations
(MICE) model that imputes missing observations. We explain the MICE model in more
detail, next.
MICE – multiple imputation chain equations
Traditional techniques that either drop missing observations and run the analysis using a
complete dataset or replace missing values with the mean or mode, are now considered
inadequate – and methodologies such as multiple imputation chain equations (MICE) have
been introduced as principled approaches to analyze incomplete data. Its main objective is
not to precisely predict the missing observations but to handle missing data in such a way
that result in a valid statistical inference. MI estimation (1) can be more efficient than
commonly-used listwise deletion (complete-cases analysis) and can correct for potential
bias; (2) it is more flexible than fully-parametric methods, e.g. maximum likelihood, purely
Bayesian analysis; and (3) since it accounts for missing-data uncertainty, it does not
underestimate the variance of estimates like single imputation methods. In brief, the model
works as follows.
The MICE model specifies a posterior density function for the missing values using
a set of predictor variables. Furthermore, it assumes that given the predictor variables used
in the multiple imputation (MI) model, the missing data would be missing at random
66
(MAR).26 In other words, after controlling for all the variables, missingness depends only
on the observed values (Azur et al. 2011). To mitigate the uncertainty associated with the
value of missing observations, the MI method makes several draws of the missing data
from their posterior predictive distribution and replaces missing values with multiple sets
of values to complete the data. Then, each complete dataset will be analyzed independently
in order to estimate the parameters of interests. Finally, parameters obtained from the m
datasets are averaged to give a single estimate with a corresponding standard error (if there
is little (much) information in the observed data to predict the missing values the
imputation results will be associated with large (small) standard errors)
We estimate the MICE model parameters in Stata using the command mi estimate. We use
route-level clustered standard errors that make our hypotheses testing more conservative
and enable us to control for unobserved route-specific factors (Eilert et al. 2017; Mccann
and Vroom 2010). Also, we follow prior studies that suggest the imputation model should
always include all variables: the dependent variable as well as any other independent,
control and auxiliary variables that may provide information about the probability of
missingness, or about the true value of the missing data and that help reducing bias and
make the MAR assumption more plausible (Azur et al. 2011 and Collins et al. 2001).
Specifically, we include in the model variables that are substantively important and are a
proxy of market attractiveness, which might be related with the propensity for incumbents
to react – and influence the number of missing values (for more discussion please see
Appendix E).
26 See the Appendix for more details on the types of missingness and how MICE works.
67
RESULTS
The results are presented in Table 2.3. When dropping all missing observations we estimate
our model using only 1764 remaining data points (Table 2.3, column - a),27 and when mean
imputing the missing IVs observations and dropping only the missing DV observations the
model is estimated using 3692 data points (Table 2.3, column - b). As already explained,
the MICE model imputes all missing observations (Table 2.3, column - c). The sign and
significance of the coefficient of interests are pretty consistent across models b and c
(models that include imputed observations). However, the results of model a (with no
imputed values) is slightly different from models with the imputed observations. We focus
on the results of the MICE model for the following reasons.
The MICE results are very much comparable to those from the series of 'hand-
imputed' models, which suggests MICE is not an obscure method that could lead to
dramatically different conclusions. Using a formal method to handle missing values in the
DV, as MICE does, however, is advisable as some differences may arise from an
econometrically sound approach to the missing values problem. And in fact, in our case,
the MICE model estimates the PE’s size effect to be positive and significant, while the
'hand-imputed' models find no effect for this variable. The results from the MICE method
are in line with a vast literature on strategy supporting the effect of firms’ size (Claussen,
Essling, and Peukert 2018).
MICE Estimation (LPM). Table 2.4 presents the results of our MICE model that
estimates how PE’s motivation and capability would influence the seriousness level of the
27 Along with 572 missing observations in DV, we also have missing observations in IVs. PE-Size has 1549
missing observations and PE-Resources has 1499 missing values. Thus, the remaining number of
observations in this scenario dropped to 1764.
68
threats posed by the PEs. 28 Notice that the model is parameterized in such a way that a
positive coefficient in the LMP regression implies a positive effect on the likelihood of
being a real threat. To measure the model fit, we calculate the overall accuracy of the
model. The accuracy is 81% which means that the model gives an accurate prediction 81%
of the time. Thus, our model improves classification accuracy by 31% compared to the
random assignment procedure with 50% accuracy.29 We start by describing the results
regarding the effects of control variables which are measured at route-, challenger- and
network-levels.
Control variables. As illustrated in Table 2.4, all control variables but one (whether
there is an incumbent’s hub in one of the two endpoint cities; p > .10) are highly significant
explaining the threat type likelihood. Market-level characteristics, whether the market is
leisure market (βLeisure = 0.125, p < .01), the distances traveled (βDistance = -0.017, p < .01 &
βDistanceSQ = -0.0007, p < .01 ) and degree of multimarket contact (βMMC = 0.85, p < .01),
significantly influence the likelihood that the posed threat by the PEs are real. These effects
could be expected from an economic point of view. For instance, the cost efficiency of
low-cost PEs compared to that of mainstream incumbents shows up more strongly on
shorter travel distances as longer routes become too costly to serve (Joskow, Werden, and
Johnson 1994). As a last control at the route level, we observe a negative association
between load factor (βLoad factor = -0.15, p < .01) and the probability of being a real threat.
This indicates that routes, where incumbents are operating efficiently, are less likely to be
28 To estimate the final model, we have used LPM. We also implemented a logistic regression and show that
the sign and significance of coefficients of interest, and therefore our overall results, remain unaffected. The
results from logistic regression not reported here but available on request.
69
PE’s targets. We also found that PEs are more likely to attack markets with a higher number
of incumbents (βNInc = 0.274, p < .01). Prior studies indicate that in the oligopolistic markets
free-rider problem is an important factor that may cause incumbent firms to underinvest in
deterrence strategies (Persson 2004; Waldman 1991). In this situation, the potential entrant
expects to encounter low or limited incumbents’ responses, thereby is more likely to attack
the market.
Results for the Hypotheses
PE’s Motivation to Entry. The results from the LPM revealed that the higher the PE’s
motivation to enter a market, it is more likely that the posed threat is a serious one. As we
discussed earlier, PE’s motivation can be reflected in market attractiveness (e.g., market
growth rate, market demand, pre-threat environment) and also, incumbent’s R&Cs. Table
2.4 illustrates that: as the incumbent’s R&Cs increases, the probability of being a real threat
decrease (βInc-resource = -.023, p < .01), which clearly indicates that when the incumbent is
capable of defending its market, the entry would be much riskier and thus the market is not
a good target for the PE. Moreover, PEs are less interested in those markets that are
important to the incumbents (βMarket-Importance = -1.5, p < .01) and will pose less serious
threats to those markets. Both market demand and market growth rate are positively
correlated with the likelihood of being a real threat (βDemand = 0.0001, p < .05 & βGrowthRate
= 0.41, p < .01). Since the market is financially attractive, the PE has a higher incentive to
attack the market, thus the posed threat would be more serious. As expected, the markets
with a higher price are more likely to be a target for the PEs (βPrice = 0.046, p < .01) thus
the threats posed on these types of routes are more likely to be real. The effect of route
70
delay is not significant indicating that two mechanisms discussed earlier might cancel each
other out, making the net impact zero.
As predicted by H2, multi-market contact between the incumbents and PE reduces
the negative impact of incumbents’ R&Cs on the likelihood of being a real threat (βInc-
resource×MMC = 0.21, p < .01). As the number of markets where the PE and the incumbents
compete with each other increases, incumbents are less likely to launch tough competitive
reactions in response to the entry, thus market entry for the PE would be less risky.
PE’s Capability of Entry. As predicted by H3 and illustrated in Table 2.4, the
probability of being a serious threat is also significantly affected by PE’s available
resources and macroeconomic factors. Specifically, PEs with higher available resources at
the under-threat market with a larger size are more likely to attack the market, thus, would
pose a more serious threat of entry (βPe resource = 0.01, p < .01; βSize = 0.01, p < .01;). Finally,
other than the market- and firm-level factors, macroeconomics factors also can influence
the probability of a threat being real. For example, when the fuel prices are higher, PEs
may slow down their network expansion, thus their posed threats are less likely to be real
(βFuelPrice = -0.17, p < .01). In sum, our results lend support for all hypotheses (H1a & b,
H2, H3).
Robustness Checks
Entry identification. To differentiate a real threat from a bluff, we relied on observed
entries. In the initial analysis, we coded entry as 1 if the PE enters a market any time after
it starts threatening the market. However, some entries may occur immediately after the
threat being established and some may be materialized long after threat establishment with
an average of six quarters after the threat. In order to assess our result sensitivity to the
71
entry definition, as a robustness check, we coded entry as 1 if the PE enters the market (a)
immediately after it starts threatening the market, (b) one quarter after threat, and (c) two,
(d) three, (e) four, (f) five, and (g) six quarters after threat. The results suggest that our key
findings are not driven by the definition of the entry (See Appendix A, Table A.2).
Reaction identification. In our initial analysis, we define incumbents’ reaction as a
10% price cut in response to the threat. To test the sensitivity of our results to this threshold,
we re-estimate our model using two alternative cut-off points, 5% price cut, and 15% price
cut. The results suggest that our key findings are not driven by the definition of reaction
either (See Appendix A, Table A.3).
Distribution of threat type. In our initial dataset (before imputation), almost 70%
of the observed threats are bluff and 30% are real. MICE keeps the distribution of observed
values when imputing the missing values. What if the final distribution of missing values
does not follow the observed distribution? To explore this question, we have done several
additional analyses. First, we assumed that all missing DV observations are real threats
(coded as 1) and mean imputed the IVs missing observations, second, we assumed that all
missing DV observations are bluff threats (coded as 0) and again mean imputed the IVs
missing observations (See Appendix A, Table A.4). We also consider that 70% of missing
DV values are bluff, then 50% is bluff, and finally 30% of missing DV values are bluff and
randomly assign “real” and “bluff” to the missing DV observations, and mean imputed the
missing IVs for each model (see Appendix A, Table A.5).30 All these different assumptions
30 We repeated this random assignment procedure of 1000 times and report the mean, and STD of the
coefficients.
72
lead to similar and consistent results, which suggest that our estimates do not depend much
on distributional assumptions.
GENERAL DISCUSSION
Contribution
Since the threat of a new entrant is one of the five competitive forces that affect a firm’s
performance (Porter 1979), it deserves ample research attention. The focus of our study is
to assess the threat posed by the potential entrant. Since incumbents do not have unlimited
resources to respond to every potential foe, threat type classification is one of the most
crucial tasks in marketing strategy (Klemz and Gruca 2003).
Several studies in economics and management have indeed explored how
incumbent firms react to the threat of entry and whether these deterrence strategies are
effective in dissuading a PE from stepping into their markets (Cookson 2017; Ethiraj and
Zhou 2019; Frésard and Valta 2016; Homburg et al. 2013; Seamans 2013). These studies,
however, implicitly assume that all threats are equal and can provoke an incumbent’s
response. In this research, we explore the more realistic and common situation in which (1)
an incumbent faces multiple threats posed by the potential entrants and (2) the nature of
competitive threats is different. A key point that we make in this study is that even when
potential entrants start threatening a market, the threat may not be a “competitive threat”
to the incumbents in the market and in that case, they would be better off by not committing
scarce resources to protect the market. As the number of potential entrants increases, the
incumbents more often misidentify the most threatening entrant (Klemz and Gruca 2003;
Yip 1982). Since incumbent does not have an unlimited budget to defend its market, a most
threatening rival would easily escape from incumbent counterattack. Our study contributes
73
to the competitive dynamic literature by empirically distinguishing real threats from the
bluff and is among the first studies that examine the extent to which some entry threats are
highly threatening (i.e., are credible or serious threats) while others are not (i.e., are bluff).
Recent studies indicate that there is heterogeneity among incumbents with regards
to market responses – with incumbents strategically increasing, decreasing or maintaining
their prior investment levels in face of a threat of entry (Frésard and Valta 2016; see also
Dafny 2005; Goolsbee and Syverson 2008; Ellison and Ellison 2011). Moreover, these
studies suggest that incumbents’ decision to respond to a threat is a very difficult task, and
one that needs to take into account various disparate factors such as whether the PE’s
product/service is a strategic substitute or complement, whether investments signal
incumbents are soft or tough defenders, and whether firms can feasibly deter entry or need
to strategically accommodate, but is silent about what comes first – the type of threat to
start with. The findings of our work can help incumbents identify the type of threat they
are facing and help them decide whether to react and in what way. For instance, if
incumbents perceive that the posed threats are a bluff, they probably should maintain the
status quo investment levels – and do nothing about those threats.
Limitations and Future Research
While this study provides novel insights into the threat type classification, it also faces
limitations that open the way to future research. The fact that the study is limited to the
airline industry implies that the results may apply in another industry somewhat differently.
However, using data from a single industry allows us to eliminate any confounding effects
from extraneous industry-specific factors, thereby improving internal validity (Eilert et al.
2017). Additional research might build on our study of threat types examinations to
74
determine the extent to which our findings are generalizable to other industries and other
contexts (i.e., product market entry).
In this research we explore the antecedent of a real threat of entry in the context of
the geographic market entry, however, firms can pose threats by introducing a new product
within the same market. In this scenario, an innovator firm has a perfect new product,
however, the firm knows that as soon as the new product hits the market, the other
incumbents would react and develop a close substitute, thus the innovator firm will be
involved in a head to head competition. Thus, in order to deceive the rivals, the innovative
firm may introduce several inferior new products in order to mislead the incumbents hoping
that the rivals would expend their limited resources on developing a close substitute to the
inferior product (Hendricks and McAfee 2006). Once the incumbent being fooled, the
entrant can introduce the superior the new product and win the market. So, another
interesting avenue for the research would be distinguishing the innovator’s inferior (bluff)
from the superior (real) new products.
There is considerable evidence that firms use ‘decoy patents’ OR failure patents to
mislead their rivals into the unprofitable research direction. For example, “in the petroleum
industry, it is common practice to patent numerous inventions, one good one in a flood of
bad inventions -- Langinier 2005 p. 522.“ Also, the pharmaceutical industry has appealing
examples that firms try to pursue this patenting “deadends” strategy to send the competitors
in wrong research directions (Hendricks and McAfee 2006; Langinier 2005). Although
there are several examples from the real world that illustrate the decoy patenting strategy,
a few studies have investigated patents as means to mislead competitors and more
interestingly, no research has distinguished a deadends (bluff) patents from a real one. So,
75
another promising avenue for the research would be exploring the characteristics of the
real patent.
Firms can also pose threats by entering a market on a small scale. In this situation,
the entrant firm intentionally invests limited resources in multiple markets where it does
not compete yet and establish a small position in those markets (Upson et al. 2012). By
doing this micro-entry strategy, firms can develop a foothold in multiple markets by
allocating a few resources. At a given point in time, the firm who owns a foothold can
attack the market on a larger scale, may withdraw the foothold. Since an incumbent cannot
react to all the micro-entries, a future study should be conducted to explore that
characteristics of a serious foothold that might lead to the actual entry in the future.
Furthermore, Clark and Montgomery (1998) indicate that an incumbent’s
willingness and ability to defend its market enhance its reputation as a “credible defender,”
and this reputation may deter potential entrants from attacking incumbents’ markets. Thus,
another interesting opportunity for future research lies in empirically investigating to what
extent an incumbent’s reputation is associated with the likelihood of the threat being real.
Since in the airline industry, incumbents usually drop prices in response to the
threat of entry, we used a level of price-cut as the main criterion to classified threats into
real and bluff. However, in other industries, incumbents may react to the threat of entry by
improving other aspects of marketing mix such as investing in their quality, advertising,
etc. An operationalization of threat classification that uses other types of reaction (or a
combination of them) would advance our current state of knowledge on differentiating a
real threat of entry from a bluff.
76
In this research, we apply a MICE method to impute the missing values. The main
assumption of the MICE method is that after controlling for some factors, the missing data
would become missing at random (MAR). However, future research could develop more
sophisticated and advanced techniques to relax this assumption and be able to treat missing
data as “Missing not at Random (MNAR).
In conclusion, despite being limited to a single industry our research highlights an
understudied area in marketing strategy by exploring the factors that can help firms to draw
a distinction between a real market threat and bluff. To achieve this goal, we took the first
step in that direction and hope our findings stimulate further interest in the study of the
market threats phenomena as a complex process involving the interactions between
incumbents and potential entrants.
77
FIGURE 2.1: Conceptual Framework
Controls
− Market Characteristics
− Incumbents’ Characteristics
− Time and firm fixed effects
Attractiveness: − Market Demand
− Market Growth Rate
− Incumbents Resource & Capabilities
− Incumbent’s Market Network Structure
− Pre-threat environment
Likelihood of Serious Threat
Capability: − PE’s Resource & Capabilities
− PE’s Size
− Fuel Price
Motivation
Capability
Multi-market
78
FIGURE 2.2: Threat Establishment
Washington
(IAD)
Miami
(MIA) IAD-MIA
DL is an incumbent
a: T=0
Denver
(DEN)
WN serves
DEN-MIA
Washington
(IAD)
Miami
(MIA) IAD-MIA
DL is an incumbent
b: T=1
Atlanta
(ATL)
WN serves
ATL-IAD WN is a potential entrant and threatens to enter IAD-MIA
Washington
(IAD)
Miami
(MIA) IAD-MIA
DL is an incumbent
Denver
(DEN)
WN serves
DEN-MIA
c: T=2
WN represents Southwest & DL represents Delta
79
TABLE 2.1a: Serious Threat vs. Bluff Threat Classification
TABLE 2.1b: Data Structure
Observed
PE Move
Entry=0 Entry=1 Total
Partially Unobserved
Type of Threat
Y=0 (Bluff) Nunobserved_bluff 0 Nbluff
Y=1 (Real) Nunobserved_real Nobserved_real Nreal
Total Nunobserved Nobserved Ntotal
Nunobserved_bluff (Nunobserved_real) refers to the unobserved number of the bluff (real) threats;
Nobserved_real refers to the observed number of real threats
Observed
PE Entry after Observed PE Threat
Entry=1 Entry=0
Observed Incumbents’
Move
Reaction Real Unobserved
No Reaction Real Bluff
80
TABLE 2.2: Descriptive Statistics and Correlation Matrix
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 Threat Type 1.00
2 Inc_resource -0.13 1.00
3 Market Growth 0.06 0.00 1.00
4 Market Demand 0.12 0.10 -0.05 1.00
5 Pre-entry Price -0.15 0.21 0.05 -0.04 1.00
6 Pre-entry OTP 0.00 0.14 0.04 0.26 0.10 1.00
7 INCRouteImportance 0.11 0.31 0.08 0.17 0.21 0.14 1.00
8 PE_resource 0.47 0.00 0.06 0.46 0.00 0.12 0.20 1.00
9 PE_size 0.43 -0.08 -0.10 -0.06 0.05 0.03 0.22 0.30 1.00
10 FuelPrice -0.08 0.04 -0.22 -0.23 0.16 -0.18 -0.05 -0.10 0.16 1.00
11 Market Distance 0.02 -0.12 0.02 0.32 0.43 0.16 -0.09 0.20 -0.02 -0.07 1.00
12 MMC -0.12 0.16 0.14 0.02 0.34 0.00 0.07 -0.02 -0.30 0.01 0.06 1.00
13 Incumbent_num 0.34 -0.09 0.00 0.56 -0.24 0.08 0.11 0.55 0.08 -0.13 0.12 -0.06 1.00
14 Leisure 0.14 -0.24 -0.07 0.05 -0.39 -0.08 -0.23 0.02 0.01 -0.09 0.01 -0.17 0.16 1.00
15 HUB -0.18 0.33 0.01 0.11 0.21 0.11 0.34 -0.03 -0.11 0.10 -0.06 0.07 -0.11 -0.23 1.00
16 Load_factor 0.19 -0.31 -0.11 0.45 -0.38 0.11 -0.16 0.28 0.07 -0.11 0.21 -0.23 0.59 0.42 -0.20 1.00
81
*p < .10, **p < .05, ***p < .01
Variance-Covariance is clustered at route level which is equivalent to route fixed effects.
TABLE 2.3: Results Across Different Imputation
Techniques
MODEL A MODEL B MODEL C
Obs: 1764 Obs: 3692 Obs: 4245
Variables Coef. P>z Coef. P>z Coef. P>t
MOTIVATION
Inc_resource -0.0272** 0.041 -0.0278*** 0.000 -0.02*** 0.00
Market Growth 0.4875*** 0.000 0.4645*** 0.000 0.415*** 0
Market Demand 0.0000 0.316 0.0001*** 0.000 0.000*** 0.01
Pre-entry Price 0.0086 0.599 0.0367*** 0.001 0.046*** 0
Pre-entry OTP 0.0011 0.916 0.0113 0.120 0.006 0.45
INC_Route_Importance -1.1002*** 0.000 -1.6260*** 0.000 -1.49*** 0
CAPABILITY
PE_Resource 0.0951*** 0.000 0.0731*** 0.000 0.098*** 0
PE_Size -0.0370** 0.015 -0.0012 0.742 0.009*** 0
FuelPrice -0.1401** 0.035 -0.1775*** 0.000 -0.17*** 0
CONTROLS
Market Distance 0.0147** 0.013 0.0233*** 0.000 0.017*** 0
DistanceSQ -0.0005** 0.032 -0.0009*** 0.000 -
0.000***
0
MMC 1.0929*** 0.000 0.6143*** 0.000 0.850*** 0
Incumbent_num 0.1326** 0.015 0.3120*** 0.000 0.274*** 0
Leisure 0.0469** 0.030 0.1135*** 0.000 0.124*** 0
Hub -0.0341 0.196 -0.0053 0.762 0.001 0.95
Load_factor -0.0950** 0.050 -0.1639*** 0.000 -0.14*** 0
INTERACTION Inc_Resource*MMC 0.1212 0.374 0.2495*** 0.000 0.215*** 0.00
82
TABLE 2.4: Result with Multiple Imputation Chained Equations
*p < .10, **p < .05, ***p < .01
Variance-Covariance is clustered at route level which is equivalent to route fixed effects.
Coef. Std. Err. t P>|t| [95% Conf. Interval] M
oti
vati
on
(Pre
-Th
rea
t P
erio
d) INC_RESOURCE -0.0229*** 0.0076 -3.04 0.003 -0.0378 -0.0081
MARKET GROWTH RATE 0.4155*** 0.0912 4.56 0 0.2365 0.5945
MARKET DEMAND 0.0001** 0 2.51 0.012 0 0.0001
MARKET PRICE LEVEL 0.0465*** 0.0128 3.62 0 0.0212 0.0718
ON TIME PERFORMANCE 0.0061 0.0082 0.75 0.454 -0.0099 0.0222
INC MARKET IMPORTANCE -1.4946*** 0.1641 -9.11 0 -1.8166 -1.1725
Ca
pa
bil
ity
PE_RESOURCE 0.098*** 0.0099 9.93 0 0.0785 0.1174
PE_SIZE 0.0095*** 0.0024 3.98 0 0.0047 0.0143
FUEL_PRICE -0.17*** 0.0276 -6.16 0 -0.2242 -0.1158
Co
ntr
ol
Va
ria
ble
s
&
Fix
ed e
ffec
ts
MKTDISTANCE 0.0173*** 0.0042 4.14 0 0.0091 0.0255
MKTDISTANCESQ -0.0007*** 0.0002 -4.46 0 -0.001 -0.0004
MMC 0.8508*** 0.1065 7.99 0 0.6418 1.0598
INCUMBENT_NUM 0.274*** 0.0369 7.42 0 0.2013 0.3467
LEISURE 0.1249*** 0.016 7.82 0 0.0935 0.1563
HUB 0.001 0.0178 0.06 0.955 -0.034 0.036
LOAD_FACTOR -0.1493*** 0.0328 -4.55 0 -0.2138 -0.0848
INC_RESOURCE×MMC 0.2153*** 0.0624 3.45 0.001 0.0927 0.3379
PE-DUMMIES Significant for Southwest and Spirit
YEAR DUMMIES Some significant
INC-DUMMIES Mostly significant
Number of Observations: 4237
83
REFERENCES
Andreassen, T. W., van Oest, R. D., & Lervik-Olsen, L. (2018). Customer inconvenience
and price compensation: a multiperiod approach to labor-automation trade-offs in
services. Journal of Service Research, 21(2), 173-183.
Aydemir, R. (2012). Threat of market entry and low-cost carrier competition. Journal of
Air Transport Management, 23, 59-62.
Azur, Melissa J., Elizabeth A. Stuart, Constantine Frangakis, and Philip J. Leaf. "Multiple
imputation by chained equations: what is it and how does it work?" International
journal of methods in psychiatric research 20, no. 1 (2011): 40-49.
Baum, J. A., & Korn, H. J. (1996). Competitive dynamics of interfirm rivalry. Academy of
Management Journal, 39(2), 255-291.
Baum, J. A., & Korn, H. J. (1999). Dynamics of dyadic competitive interaction. Strategic
Management Journal, 20(3), 251-278.
Bayus, B. L., & Agarwal, R. (2007). The role of pre-entry experience, entry timing, and
product technology strategies in explaining firm survival. Management
Science, 53(12), 1887-1902.
Bendinelli, W. E., Bettini, H. F., & Oliveira, A. V. (2016). Airline delays, congestion
internalization and non-price spillover effects of low-cost carrier
entry. Transportation Research Part A: Policy and Practice, 85, 39-52.
Bercovitz, J., & Mitchell, W. (2007). When is more better? The impact of business scale
and scope on long-term business survival, while controlling for
profitability. Strategic Management Journal, 28(1), 61-79.
Berry, L. L., Seiders, K., & Grewal, D. (2002). Understanding service
convenience. Journal of Marketing, 66(3), 1-17.
Berry, S., & Jia, P. (2010). Tracing the woes: An empirical analysis of the airline
industry. American Economic Journal: Microeconomics, 2(3), 1-43.
Bertrand, A., Legrand, C., Carroll, R. J., De Meester, C., & Van Keilegom, I. (2017).
Inference in a survival cure model with mismeasured covariates using a simulation-
extrapolation approach. Biometrika, 104(1), 31-50.
84
Bharadwaj, S. G., Varadarajan, P. R., & Fahy, J. (1993). Sustainable competitive advantage
in service industries: a conceptual model and research propositions. Journal of
Marketing, 57(4), 83-99.
Boeker, W., Goodstein, J., Stephan, J., & Murmann, J. P. (1997). Competition in a
multimarket environment: The case of market exit. Organization Science, 8(2),
126-142.
Boguslaski, C., Ito, H., & Lee, D. (2004). Entry patterns in the southwest airlines route
system. Review of Industrial Organization, 25(3), 317-350.
Bowman, D., & Gatignon, H. (1995). Determinants of competitor response time to a new
product introduction. Journal of Marketing Research, 32(1), 42-53.
Brueckner, J. K., Lee, D., & Singer, E. S. (2013). Airline competition and domestic US
airfares: A comprehensive reappraisal. Economics of Transportation, 2(1), 1-17.
Brueckner, J. K., Lee, D., & Singer, E. (2014). City-pairs versus airport-pairs: a market-
definition methodology for the airline industry. Review of Industrial
Organization, 44(1), 1-25.
Cabral, L. M., & Ross, T. W. (2008). Are sunk costs a barrier to entry?. Journal of
Economics & Management Strategy, 17(1), 97-112.
Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: methods and applications.
Cambridge university press.
Chadwick, C., Guthrie, J. P., & Xing, X. (2016). The HR executive effect on firm
performance and survival. Strategic Management Journal, 37(11), 2346-2361.
Chandrasekaran, D., & Tellis, G. J. (2011). Getting a grip on the saddle: Chasms or
cycles?. Journal of Marketing, 75(4), 21-34.
Chen, M. J. (1996). Competitor analysis and interfirm rivalry: Toward a theoretical
integration. Academy of Management Review, 21(1), 100-134.
Chen, M. J., Kuo-Hsien, S. U., & Tsai, W. (2007). Competitive tension: The awareness-
motivation-capability perspective. Academy of Management Journal, 50(1), 101-
118.
Chen, M. J., & MacMillan, I. C. (1992). Nonresponse and delayed response to competitive
moves: The roles of competitor dependence and action irreversibility. Academy of
Management Journal, 35(3), 539-570.
Chen MJ & Miller D. (2012). Competitive dynamics: themes trends and a prospective
research platform. Academy of Management Annals 6: 1-76.
85
Chen, M. J., Smith, K. G., & Grimm, C. M. (1992). Action characteristics as predictors of
competitive responses. Management Science, 38(3), 439-455.
Chen, M. J., Venkataraman, S., Sloan Black, S., & MacMillan, I. C. (2002). The role of
irreversibilities in competitive interaction: Behavioral considerations from
organization theory. Managerial and Decision Economics, 23(4‐5), 187-207.
Cho, Y. K. (2014). Service quality and price perceptions by internet retail customers:
linking the three stages of service interaction. Journal of Service Research, 17(4),
432-445.
Clark, B. H., & Montgomery, D. B. (1998). Deterrence, reputations, and competitive
cognition. Management Science, 44(1), 62-82.
Claussen, J., Essling, C., & Peukert, C. (2018). Demand variation, strategic flexibility and
market entry: Evidence from the US airline industry. Strategic Management
Journal, 39(11), 2877-2898.
Colwell, S. R., Aung, M., Kanetkar, V., & Holden, A. L. (2008). Toward a measure of
service convenience: multiple-item scale development and empirical test. Journal
of Services Marketing, 22(2), 160-169.
Collier, J. E., & Sherrell, D. L. (2010). Examining the influence of control and convenience
in a self-service setting. Journal of the Academy of Marketing Science, 38(4), 490-
509.
Collier, J. E., & Kimes, S. E. (2013). Only if it is convenient: understanding how
convenience influences self-service technology evaluation. Journal of Service
Research, 16(1), 39-51.
Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of inclusive and
restrictive strategies in modern missing data procedures. Psychological
methods, 6(4), 330.
Connelly, B. L., Certo, S. T., Ireland, R. D., & Reutzel, C. R. (2011). Signaling theory: A
review and assessment. Journal of Management, 37(1), 39-67.
Cookson, J. A. (2017). Anticipated entry and entry deterrence: Evidence from the
American casino industry. Management Science.
Dafny, L. S. (2005). Games hospitals play: Entry deterrence in hospital procedure
markets. Journal of Economics & Management Strategy, 14(3), 513-542.
Dawes, J., Meyer-Waarden, L., & Driesener, C. (2015). Has brand loyalty declined? A
longitudinal analysis of repeat purchase behavior in the UK and the USA. Journal
of Business Research, 68(2), 425-432.
86
Day, G. S., & Wensley, R. (1983). Marketing theory with a strategic orientation. Journal
of Marketing, 47(4), 79-89.
Dirick, L., Bellotti, T., Claeskens, G., & Baesens, B. (2019). Macro-economic factors in
credit risk calculations: including time-varying covariates in mixture cure
models. Journal of Business & Economic Statistics, 37(1), 40-53.
Dixit, A. (1989). Entry and exit decisions under uncertainty. Journal of Political
Economy, 97(3), 620-638.
Dixit, A., & Chintagunta, P. K. (2007). Learning and exit behavior of new entrant discount
airlines from city-pair markets. Journal of Marketing, 71(2), 150-168.
Dunn, A. (2008). Do low-quality products affect high-quality entry? Multiproduct firms
and nonstop entry in airline markets. International Journal of Industrial
Organization, 26(5), 1074-1089.
Dunne, T., Klimek, S. D., Roberts, M. J., & Xu, D. Y. (2013). Entry, exit, and the
determinants of market structure. The RAND Journal of Economics, 44(3), 462-
487.
Eilert, M., Jayachandran, S., Kalaignanam, K., & Swartz, T. A. (2017). Does it pay to recall
your product early? An empirical investigation in the automobile industry. Journal
of Marketing, 81(3), 111-129.
Elfenbein, D. W., & Knott, A. M. (2015). Time to exit: rational, behavioral, and
organizational delays. Strategic Management Journal, 36(7), 957-975.
Eliashberg, J. and T. Robertson & T. Rymon (1996), "Market Signaling and Competitive
Bluffing: An Empirical Study," Marketing Science Institute Working Paper No. 96-
102. Cambridge, MA: Marketing Science Institute.
Ellison, G., & Ellison, S. F. (2011). Strategic entry deterrence and the behavior of
pharmaceutical incumbents prior to patent expiration. American Economic
Journal: Microeconomics, 3(1), 1-36.
Erdem, T., & Keane, M. P. (1996). Decision-making under uncertainty: Capturing dynamic
brand choice processes in turbulent consumer goods markets. Marketing
Science, 15(1), 1-20.
Ethiraj, S., & Zhou, Y. M. (2019). Fight or flight? Market positions, submarket
interdependencies, and strategic responses to entry threats. Strategic Management
Journal.
Falk, T., Hammerschmidt, M., & Schepers, J. J. (2010). The service quality-satisfaction
link revisited: exploring asymmetries and dynamics. Journal of the Academy of
Marketing Science, 38(3), 288-302.
87
Farquhar, J. D., & Rowley, J. (2009). Convenience: a services perspective. Marketing
Theory, 9(4), 425-438.
Frésard, L., & Valta, P. (2016). How does corporate investment respond to increased entry
threat? The Review of Corporate Finance Studies, 5(1), 1-35. Franco, A. M., Sarkar, M. B., Agarwal, R., & Echambadi, R. (2009). Swift and smart: The
moderating effects of technological capabilities on the market pioneering–firm
survival relationship. Management Science, 55(11), 1842-1860.
García-Fernández, J., Gálvez-Ruíz, P., Fernández-Gavira, J., Vélez-Colón, L., Pitts, B., &
Bernal-García, A. (2018). The effects of service convenience and perceived quality
on perceived value, satisfaction and loyalty in low-cost fitness centers. Sport
Management Review, 21(3), 250-262.
Gatignon, H., Anderson, E., & Helsen, K. (1989). Competitive reactions to market entry:
Explaining interfirm differences. Journal of Marketing Research, 26(1), 44-55.
Gatignon, H., Robertson, T. S., & Fein, A. J. (1997). Incumbent defense strategies against
new product entry. International Journal of Research in Marketing, 14(2), 163-
176.
Gayle, P. G., & Wu, C. Y. (2013). A re-examination of incumbents’ response to the threat
of entry: Evidence from the airline industry. Economics of Transportation, 2(4),
119-130.
Gerardi, K. S., & Shapiro, A. H. (2009). Does competition reduce price dispersion? New
evidence from the airline industry. Journal of Political Economy, 117(1), 1-37.
Geroski, P. A. (1995). What do we know about entry?. International Journal of Industrial
Organization, 13(4), 421-440.
Geroski, P. A., Mata, J., & Portugal, P. (2010). Founding conditions and the survival of
new firms. Strategic Management Journal, 31(5), 510-529.
Ghemawat, P. (2016). Evolving ideas about business strategy. Business History Review,
90(4), 727-749.
Gimeno, J. (1999). Reciprocal threats in multimarket rivalry: Staking out ‘spheres of
influence’ in the US airline industry. Strategic Management Journal, 20(2), 101-
128.
Goetz, C. F., & Shapiro, A. H. (2012). Strategic alliance as a response to the threat of entry:
Evidence from airline codesharing. International Journal of Industrial
Organization, 30(6), 735-747.
88
Goldhaber, D., Krieg, J., & Theobald, R. (2014). Knocking on the door to the teaching
profession? Modeling the entry of prospective teachers into the
workforce. Economics of Education Review, 43, 106-124.
Goolsbee, A., & Syverson, C. (2008). How do incumbents respond to the threat of entry?
Evidence from the major airlines. The Quarterly Journal of Economics, 123(4),
1611-1633.
Greenfield, D. (2014). Competition and service quality: New evidence from the airline
industry. Economics of Transportation, 3(1), 80-89.
Grewal, R., Chandrashekaran, M., & Citrin, A. V. (2010). Customer satisfaction
heterogeneity and shareholder value. Journal of Marketing Research, 47(4), 612-
626.
Groening, C., Mittal, V., & “Anthea” Zhang, Y. (2016). Cross-validation of customer and
employee signals and firm valuation. Journal of Marketing Research, 53(1), 61-76.
Guidice, R. M., Alder, G. S., & Phelan, S. E. (2009). Competitive bluffing: An examination
of a common practice and its relationship with performance. Journal of Business
Ethics, 87(4), 535-553.
Guiltinan, J. P., & Gundlach, G. T. (1996). Aggressive and predatory pricing: A framework
for analysis. Journal of Marketing, 60(3), 87-102.
Haenlein, M., Kaplan, A. M., & Schoder, D. (2006). Valuing the real option of abandoning
unprofitable customers when calculating customer lifetime value. Journal of
Marketing, 70(3), 5-20.
Hambrick, D. C., Cho, T. S., & Chen, M. J. (1996). The influence of top management team
heterogeneity on firms' competitive moves. Administrative Science Quarterly, 659-
684.
Hauser, J. R., & Shugan, S. M. (2008). Defensive marketing strategies. Marketing
Science, 27(1), 88-110.
Helfat, C. E., & Lieberman, M. B. (2002). The birth of capabilities: market entry and the
importance of pre‐history. Industrial and Corporate Change, 11(4), 725-760.
Hellevik, O. (2009). Linear versus logistic regression when the dependent variable is a
dichotomy. Quality & Quantity, 43(1), 59-74.
Hendricks, K., & McAfee, R. P. (2006). Feints. Journal of Economics & Management
Strategy, 15(2), 431-456.
Hitsch, G. J. (2006). An empirical model of optimal dynamic product launch and exit under
demand uncertainty. Marketing Science, 25(1), 25-50.
89
Homburg, C., Koschate, N., & Hoyer, W. D. (2005). Do satisfied customers really pay
more? A study of the relationship between customer satisfaction and willingness to
pay. Journal of Marketing, 69(2), 84-96.
Homburg, C., Fürst, A., Ehrmann, T., & Scheinker, E. (2013). Incumbents’ defense
strategies: a comparison of deterrence and shakeout strategy based on evolutionary
game theory. Journal of the Academy of Marketing Science, 41(2), 185-205.
Horn, J. T., Lovallo, D. P., & Viguerie, S. P. (2005). Beating the odds in market entry. The
McKinsey Quarterly, 4, 34-45.
Hsieh, K. Y., Tsai, W., & Chen, M. J. (2015). If they can do it, why not us? Competitors
as reference points for justifying escalation of commitment. Academy of
Management Journal, 58(1), 38-58.
Huse, C., & Oliveira, A. V. (2012). Does product differentiation soften price reactions to
entry? Evidence from the airline industry. Journal of Transport Economics and
Policy (JTEP), 46(2), 189-204.
Ito, H., & Lee, D. (2003). Low cost carrier growth in the US airline industry: past, present,
and future. Brown University Department of Economics Paper, (2003-12).
Jaggia, S. (2011). Identifiability of the misspecified split hazard models. Applied
Economics, 43(24), 3441-3447.
Jayachandran, S., Gimeno, J., & Varadarajan, P. R. (1999). The theory of multimarket
competition: A synthesis and implications for marketing strategy. Journal of
Marketing, 63(3), 49-66.
Johnson, J., & Tellis, G. J. (2008). Drivers of success for market entry into China and
India. Journal of Marketing, 72(3), 1-13.
Joskow, A. S., Werden, G. J., & Johnson, R. L. (1994). Entry, exit, and performance in
airline markets. International Journal of Industrial Organization, 12(4), 457-471.
Kalra, A., Rajiv, S., & Srinivasan, K. (1998). Response to competitive entry: A rationale
for delayed defensive reaction. Marketing Science, 17(4), 380-405.
Karaer, Ö., & Erhun, F. (2015). Quality and entry deterrence. European Journal of
Operational Research, 240(1), 292-303.
Ketokivi, M., & McIntosh, C. N. (2017). Addressing the endogeneity dilemma in
operations management research: Theoretical, empirical, and pragmatic
considerations. Journal of Operations Management, 52, 1-14.
Kirmani, A., & Rao, A. R. (2000). No pain, no gain: A critical review of the literature on
signaling unobservable product quality. Journal of Marketing, 64(2), 66-79.
90
Klein, J. P., Van Houwelingen, H. C., Ibrahim, J. G., & Scheike, T. H. (Eds.).
(2016). Handbook of Survival Analysis. CRC Press.
Klemz, B. R., & Gruca, T. S. (2003). Dueling or the battle royale? The impact of task
complexity on the evaluation of entry threat. Psychology & Marketing, 20(11), 999-
1016.
Kuester, S., Homburg, C., & Robertson, T. S. (1999). Retaliatory behavior to new product
entry. Journal of Marketing, 63(4), 90-106.
Kumar, N. (2006). Strategies to fight low-cost rivals. Harvard Business Review, 84(12),
104.
Lamey, L. (2014). Hard economic times: a dream for discounters. European Journal of
Marketing, 48(3/4), 641-656.
Langinier, C. (2005). Using patents to mislead rivals. Canadian Journal of
Economics/Revue canadienne d'economique, 38(2), 520-545
Lieberman, M. B., Lee, G. K., & Folta, T. B. (2017). Entry, exit, and the potential for
resource redeployment. Strategic Management Journal, 38(3), 526-544.
Luoma, J., Falk, T., Totzek, D., Tikkanen, H., & Mrozek, A. (2018). Big splash, no waves?
C ognitive mechanisms driving incumbent firms’ responses to low‐price market
entry strategies. Strategic Management Journal, 39(5), 1388-1410.
Lurkin, V., Garrow, L. A., Higgins, M. J., Newman, J. P., & Schyns, M. (2017).
Accounting for price endogeneity in airline itinerary choice models: An application
to Continental US markets. Transportation Research Part A: Policy and
Practice, 100, 228-246.
Lusch, R. F., Vargo, S. L., & O’brien, M. (2007). Competing through service: Insights
from service-dominant logic. Journal of Retailing, 83(1), 5-18.
Marcel, J. J., Barr, P. S., & Duhaime, I. M. (2011). The influence of executive cognition
on competitive dynamics. Strategic Management Journal, 32(2), 115-138.
Mayer, C., & Sinai, T. (2003). Network effects, congestion externalities, and air traffic
delays: Or why not all delays are evil. American Economic Review, 93(4), 1194-
1215.
McCann, B. T., & Vroom, G. (2010). Pricing response to entry and agglomeration
effects. Strategic Management Journal, 31(3), 284-305.
McGrath, R. G., Chen, M. J., & MacMillan, I. C. (1998). Multimarket maneuvering in
uncertain spheres of influence: Resource diversion strategies. Academy of
Management Review, 23(4), 724-740.
91
Min, S., Kalwani, M. U., & Robinson, W. T. (2006). Market pioneer and early follower
survival risks: A contingency analysis of really new versus incrementally new
product-markets. Journal of Marketing, 70(1), 15-33.
Montgomery, D. B., Moore, M. C., & Urbany, J. E. (2005). Reasoning about competitive
reactions: Evidence from executives. Marketing Science, 24(1), 138-149.
Mumbower, S., Garrow, L. A., & Higgins, M. J. (2014). Estimating flight-level price
elasticities using online airline data: A first step toward integrating pricing, demand,
and revenue optimization. Transportation Research Part A: Policy and
Practice, 66, 196-212.
Nikolaeva, R. (2007). The dynamic nature of survival determinants in e-
commerce. Journal of the Academy of Marketing Science, 35(4), 560-571.
O'Brien, J., & Folta, T. (2009). Sunk costs, uncertainty and market exit: A real options
perspective. Industrial and Corporate Change, 18(5), 807-833.
Obeng, E., Luchs, R., Inman, J. J., & Hulland, J. (2016). Survival of the fittest: How
competitive service overlap and retail format impact incumbents’ vulnerability to
new entrants. Journal of Retailing, 92(4), 383-396.
Oliveira, A. V., & Huse, C. (2009). Localized competitive advantage and price reactions
to entry: Full-service vs. low-cost airlines in recently liberalized emerging
markets. Transportation Research Part E: Logistics and Transportation
Review, 45(2), 307-320.
Papyrina, V. (2007). When, how, and with what success? The joint effect of entry timing
and entry mode on survival of Japanese subsidiaries in China. Journal of
International Marketing, 15(3), 73-95.
Panagopoulos, N. G., Mullins, R., & Avramidis, P. (2018). Sales Force Downsizing and
Firm-Idiosyncratic Risk: The Contingent Role of Investors’ Screening and Firm’s
Signaling Processes. Journal of Marketing, 82(6), 71-88.
Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). Servqual: A multiple-item scale
for measuring consumer perc. Journal of Retailing, 64(1), 12.
Parasuraman, A., & Grewal, D. (2000). The impact of technology on the quality-value-
loyalty chain: a research agenda. Journal of the Academy of Marketing
Science, 28(1), 168-174.
Parise, G. (2018). Threat of entry and debt maturity: Evidence from airlines. Journal of
Financial Economics, 127(2), 226-247.
Pauwels, K., & D’Aveni, R. (2016). The formation, evolution and replacement of price–
quality relationships. Journal of the Academy of Marketing Science, 44(1), 46-65.
92
Pe'er, A., Vertinsky, I., & Keil, T. (2016). Growth and survival: The moderating effects of
local agglomeration and local market structure. Strategic Management
Journal, 37(3), 541-564.
Persson, L. (2004). Predation and mergers: Is merger law counterproductive?. European
Economic Review, 48(2), 239-258.
Pirkul, H., & Schilling, D. A. (1998). An efficient procedure for designing single allocation
hub and spoke systems. Management Science, 44(12-part-2), S235-S242.
Porter, M. E. (1980). Competitive strategy: Techniques for analyzing industries and
competitors. Simon and Schuster.
Porter, M. E. (1985). Competitive advantage: Creating and sustaining superior
performance. Simon and Schuster.
Prabhu, J., & Stewart, D. W. (2001). Signaling strategies in competitive interaction:
Building reputations and hiding the truth. Journal of Marketing Research, 38(1),
62-72.
Prasad Mishra, D., & Bhabra, H. S. (2001). Assessing the economic worth of new product
pre-announcement signals: Theory and empirical evidence. Journal of Product &
Brand Management, 10(2), 75-93
Prince, J. T., & Simon, D. H. (2009). Multimarket contact and service quality: Evidence
from on-time performance in the US airline industry. Academy of Management
Journal, 52(2), 336-354.
Prince, J. T., & Simon, D. H. (2014). Do incumbents improve service quality in response
to entry? Evidence from airlines' on-time performance. Management
Science, 61(2), 372-390.
Prins, R., & Verhoef, P. C. (2007). Marketing communication drivers of adoption timing
of a new e-service among existing customers. Journal of Marketing, 71(2), 169-
183.
Reibstein, D. J., & Wittink, D. R. (2005). Competitive responsiveness. Marketing Science,
24 (Winter), 8-11.
Risselada, H., Verhoef, P. C., & Bijmolt, T. H. (2014). Dynamic effects of social influence
and direct marketing on the adoption of high-technology products. Journal of
Marketing, 78(2), 52-68.
Robertson, T. S., Eliashberg, J., & Rymon, T. (1995). New product announcement signals
and incumbent reactions. Journal of Marketing, 59(3), 1-15.
Robinson, W. T. (1988). Marketing mix reactions to entry. Marketing Science, 7(4), 368-
385.
93
Robinson, W. T., & Min, S. (2002). Is the first to market the first to fail? Empirical evidence
for industrial goods businesses. Journal of Marketing Research, 39(1), 120-128.
Ryans, A. (2009). Beating low cost competition: How premium brands can respond to cut-
price rivals. John Wiley & Sons.
Rust, R. T., Lemon, K. N., & Zeithaml, V. A. (2004). Return on marketing: Using customer
equity to focus marketing strategy. Journal of Marketing, 68(1), 109-127.
Seamans, R. C. (2013). Threat of entry, asymmetric information, and pricing. Strategic
Management Journal, 34(4), 426-444.
Seiders, K., Voss, G. B., Godfrey, A. L., & Grewal, D. (2007). SERVCON: development
and validation of a multidimensional service convenience scale. Journal of the
Academy of Marketing Science, 35(1), 144-156.
Sengupta, A., & Wiggins, S. N. (2014). Airline pricing, price dispersion, and ticket
characteristics on and off the internet. American Economic Journal: Economic
Policy, 6(1), 272-307.
Shankar, V. (1997). Pioneers' marketing mix reactions to entry in different competitive
game structures: Theoretical analysis and empirical illustration. Marketing
Science, 16(3), 271-293.
Shaw, S. (2016). Airline marketing and management. 6th ed.). Aldershot: Ashgate
Publishing Ltd.
Simon, D. (2005). Incumbent pricing responses to entry. Strategic Management
Journal, 26(13), 1229-1248.
Sinha, R. K., & Chandrashekaran, M. (1992). A split hazard model for analyzing the
diffusion of innovations. Journal of Marketing Research, 29(1), 116-127.
Sinha, R. K., & Noble, C. H. (2008). The adoption of radical manufacturing technologies
and firm survival. Strategic Management Journal, 29(9), 943-962.
Sivakumar, K., Li, M., & Dong, B. (2014). Service quality: The impact of frequency,
timing, proximity, and sequence of failures and delights. Journal of
Marketing, 78(1), 41-58.
Sousa, C. M., & Tan, Q. (2015). Exit from a foreign market: do poor performance, strategic
fit, cultural distance, and international experience matter? Journal of International
Marketing, 23(4), 84-104.
Srinivasan, R., Lilien, G. L., & Rangaswamy, A. (2004). First in, first out? The effects of
network externalities on pioneer survival. Journal of Marketing, 68(1), 41-58.
94
Srivastava, R. K., Fahey, L., & Christensen, H. K. (2001). The resource-based view and
marketing: The role of market-based assets in gaining competitive
advantage. Journal of Management, 27(6), 777-802.
Steen, E. V. D. (2016). A formal theory of strategy. Management Science, 63(8), 2616-
2636.
Steenkamp, J. B. E., Nijs, V. R., Hanssens, D. M., & Dekimpe, M. G. (2005). Competitive
reactions to advertising and promotion attacks. Marketing Science, 24(1), 35-54.
Sunny Yang, S. J., & Emma Liu, Y. (2015). Anticipated responses: The positive side of
elicited reactions to competitive action. Journal of the Operational Research
Society, 66(2), 316-330.
Talay, M. B., Akdeniz, M. B., & Kirca, A. H. (2017). When do the stock market returns to
new product preannouncements predict product performance? Empirical evidence
from the US automotive industry. Journal of the Academy of Marketing
Science, 45(4), 513-533.
Terza, J. V., Basu, A., & Rathouz, P. J. (2008). Two-stage residual inclusion estimation:
addressing endogeneity in health econometric modeling. Journal of Health
Economics, 27(3), 531-543.
Tsai, W., Su, K. H., & Chen, M. J. (2011). Seeing through the eyes of a rival: Competitor
acumen based on rival-centric perceptions. Academy of Management
Journal, 54(4), 761-778.
Tzu, S. 1963. The art of war. (Translated by S. B Griffith.) Oxford: Oxford University
Press
Ngoc Thuy, P. (2011). Using service convenience to reduce perceived cost. Marketing
Intelligence & Planning, 29(5), 473-487.
Trigeorgis, L., & Reuer, J. J. (2017). Real options theory in strategic
management. Strategic Management Journal, 38(1), 42-63.
Umashankar, N., Bhagwat, Y., & Kumar, V. (2017). Do loyal customers really pay more
for services?. Journal of the Academy of Marketing Science, 45(6), 807-826.
Upson, J. W., Ketchen Jr, D. J., Connelly, B. L., & Ranft, A. L. (2012). Competitor analysis
and foothold moves. Academy of Management Journal, 55(1), 93-110.
Van Kranenburg, H. L., Palm, F. C., & Pfann, G. A. (2002). Exit and survival in a
concentrating industry: The case of daily newspapers in the Netherlands. Review of
Industrial Organization, 21(3), 283-303.
95
Voss, G. B., Godfrey, A., & Seiders, K. (2010). How complementarity and substitution
alter the customer satisfaction–repurchase link. Journal of Marketing, 74(6), 111-
127.
Waldman, M. (1991). The role of multiple potential entrants/sequential entry in
noncooperative entry deterrence. The Rand Journal of Economics, 446-453.
Wang, Q., Chen, Y., & Xie, J. (2010). Survival in markets with network effects: product
compatibility and order-of-entry effects. Journal of Marketing, 74(4), 1-14.
Wang, H., Gurnani, H., & Erkoc, M. (2016). Entry deterrence of capacitated competition
using price and non‐price strategies. Production and Operations Management,
25(4), 719-735.
Wang, R. D., & Shaver, J. M. (2014). Competition-driven repositioning. Strategic
Management Journal, 35(11), 1585-1604.
Wang, Q., & Xie, J. (2011). Will consumers be willing to pay more when your competitors
adopt your technology? The impacts of the supporting-firm base in markets with
network effects. Journal of Marketing, 75(5), 1-17.
Wei, W., & Hansen, M. (2003). Cost economics of aircraft size. Journal of Transport
Economics and Policy (JTEP), 37(2), 279-296.
Wei, W., & Hansen, M. (2005). Impact of aircraft size and seat availability on airlines’
demand and market share in duopoly markets. Transportation Research Part E:
Logistics and Transportation Review, 41(4), 315-327.
Wieseke, J., Alavi, S., & Habel, J. (2014). Willing to pay more, eager to pay less: The role
of customer loyalty in price negotiations. Journal of Marketing, 78(6), 17-37.
Yip, G. S. (1982). Diversification entry: Internal development versus acquisition. Strategic
Management Journal, 3(4), 331-345.
Zajac, E. J., & Bazerman, M. H. (1991). Blind spots in industry and competitor analysis:
Implications of interfirm (mis) perceptions for strategic decisions. Academy of
Management Review, 16(1), 37-56.
Zeithaml, V. A. (1988). Consumer perceptions of price, quality, and value: a means-end
model and synthesis of evidence. Journal of Marketing, 52(3), 2-22.
Zeithaml, V. A., Bitner, M. J., Gremler, D. D., & Pandit, A. (2006). Services marketing:
Integrating customer focus across the firm.
96
APPENDIX A: DESCRIPTIVE STATISTICS AND ROBUSTNESS CHECKS
FIGURE A.1: Incumbents’ Marketing-Mix Tactics Before and After a Challenger’s Entry
97
TABLE A.1:Descriptive Statistics and Correlation Matrix
Notes: Bold: p < .05. TTE=Time to Exit, PE-PR = Pre-entry Price, PE-FR=Pre-entry frequency, PE-PK=Pre-entry peak frequency, PE-OTP=Pre-entry
OTP, PE-PS=Pre-entry plane size, INC-PR= Incumbent post entry price-cut, INC-FR= Incumbent post entry frequency, INC-PK=Incumbent post entry
peak frequency, INC-OTP=Incumbent post entry OTP, CH-PR=Challenger, F-PR = Fuel Price, CH-Size= Challenger Size, CH-IMP= Challenger route
importance, IN-IMP= Incumbent route importance, NofINC= Number of Incumbents.
a: Mean of time-to-exit is calculated among exit observations only.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
TTEa 1
PE-PR -
0.147 1
PE-FR 0.027 -
0.069 1
PE-PK -
0.143 0.061 -0.02 1
PE-OTP -
0.451 0.088 -0.16 0.016 1
PE-PS 0.088 0.149 0.288 -0.07 -
0.182 1
INC-PR -
0.036 0.402 0.073
-
0.016
-
0.059
-
0.129 1
INC-FR 0.011 0.023 -
0.196 -0.08 0.089 0.121
-
0.046 1
INC-PK 0.083 0.056 -
0.029 0.484
-
0.084
-
0.043 0.126 0.007 1
INC-
OTP
-
0.158 0.051
-
0.095 0.041 0.343
-
0.103
-
0.073
-
0.045
-
0.073 1
CH-PR 0.187 0.374 -
0.271 0.018
-
0.041 0.182
-
0.291 0.088
-
0.001
-
0.018 1
Hub -
0.026 0.111 0.176 0.032
-
0.099 0.051 0.126
-
0.014 0.032 0.017
-
0.077 1
2ND entry -
0.001 0.004 0.008
-
0.014
-
0.001
-
0.016 0.011
-
0.005
-
0.004
-
0.005
-
0.009
-
0.013 1
Distance 0.036 0.538 -0.36 0.003 0.028 0.347 -
0.136 0.154 0.004 0.039 0.655
-
0.115
-
0.018 1
MMC 0.148 0.156 0.104 -
0.146
-
0.246
-
0.028 0.245 0.01
-
0.011 0.165
-
0.123 0.171
-
0.005
-
0.133 1
F-PR 0.029 0.015 -
0.065 0.193 0.139
-
0.035
-
0.222
-
0.055
-
0.076
-
0.099 0.21
-
0.003 0.007 0.043
-
0.318 1
CH-Size -
0.072 -0.08
-
0.128 0.217 0.178
-
0.109
-
0.201
-
0.067 -0.03
-
0.186 0.271 0.005
-
0.038 0.048
-
0.559 0.256 1
Demand 0.112 0.057 0.353 -.09 -
0.057 0.443
-
0.078 0.129 0.090 0.053 0.091 0.128 0.010 0.254
-
0.126 0.007
-
0.095 1
CH-IMP 0.476 0.113 0.081 -
0.176
-
0.349
-
0.024 0.276 0.000
-
0.071 0.169
-
0.004 0.071 0.036
-
0.069 0.728
-
0.173
-
0.581 0.008 1
IN-IMP -
0.096 0.061 0.505
-
0.052
-
0.141 0.284 0.06
-
0.008 0.086 0.042
-
0.182 0.388 0.005
-
0.253 0.223
-
0.144
-
0.072 0.154 0.098 1
NofINC 0.046 -
0.057 0.187
-
0.074 0.005 0.140 0.039
-
0.009 0.069
-
0.031
-
0.031
-
0.022 0.075 0.028
-
0.176 0.016
-
0.152 0.512 0.024
-
0.026 1
98
TABLE A.2: Robustness Check – Entry Definition31
31 We do not report year and firm fixed effects in this table; *p < .10, **p < .05, ***p < .01
A B C D E F G
Missing DV: 719
obs
Missing DV:
703 obs
Missing DV:
687 obs
Missing DV:
673 obs
Missing DV:
671 obs
Missing DV:
663 obs
Missing DV:
658 obs
Variables Coef. P>z Coef. P>z Coef. P>z Coef. P>z Coef. P>z Coef. P>z Coef. P>z
MOTIVATION
Inc_resource -0.013 0.002 -0.020 0.001 -0.027 0.000 -0.027 0.000 -0.024 0.002 -0.023 0.003 -0.026 0.001
Market Growth 0.401 0.000 0.442 0.000 0.318 0.000 0.308 0.002 0.374 0.000 0.392 0.000 0.373 0.000
Market Demand 0.000 0.009 0.000 0.222 0.000 0.240 0.000 0.828 0.000 0.739 0.000 0.694 0.000 0.394
Pre-threat Price 0.073 0.000 0.061 0.000 0.066 0.000 0.066 0.000 0.067 0.000 0.061 0.000 0.059 0.000
Pre-threat Delay 0.010 0.058 0.015 0.022 0.012 0.078 0.016 0.041 0.018 0.024 0.015 0.057 0.014 0.075
INC_Route_Importance -1.014 0.000 -0.988 0.000 -0.969 0.000 -1.088 0.000 -1.097 0.000 -1.103 0.000 -1.192 0.000
CAPABILITY
PE_Resource 0.066 0.000 0.083 0.000 0.094 0.000 0.082 0.000 0.088 0.000 0.091 0.000 0.091 0.000
PE_Size 0.000 0.915 0.003 0.169 0.004 0.033 0.006 0.012 0.004 0.061 0.004 0.024 0.005 0.025
FuelPrice -0.105 0.000 -0.118 0.000 -0.131 0.000 -0.135 0.000 -0.149 0.000 -0.125 0.000 -0.125 0.000
CONTROLS
Market Distance 0.001 0.617 0.008 0.012 0.009 0.013 0.012 0.003 0.012 0.002 0.013 0.001 0.014 0.000
DistanceSQ 0.000 0.043 -0.001 0.000 -0.001 0.001 -0.001 0.000 -0.001 0.000 -0.001 0.000 -0.001 0.000
MMC 0.592 0.000 0.684 0.000 0.726 0.000 0.808 0.000 0.748 0.000 0.887 0.000 0.901 0.000
Incumbent_num 0.449 0.000 0.445 0.000 0.415 0.000 0.409 0.000 0.391 0.000 0.390 0.000 0.366 0.000
Leisure 0.051 0.000 0.046 0.000 0.068 0.000 0.063 0.000 0.068 0.000 0.070 0.000 0.082 0.000
Hub 0.043 0.001 0.019 0.197 0.015 0.329 0.016 0.307 0.020 0.236 -0.001 0.960 0.005 0.787
Load_factor -0.347 0.000 -0.290 0.000 -0.275 0.000 -0.283 0.000 -0.248 0.000 -0.251 0.000 -0.237 0.000
INTERACTION Inc Resource*MMC 0.146 0.000 0.294 0.000 0.307 0.000 0.310 0.000 0.297 0.000 0.282 0.000 0.291 0.000
99
TABLE A.3: Robustness Check-Reaction Definition
10% PRICE CUT 5% PRICE CUT 15% PRICE CUT
MISSING DV: 572 OBS MISSING DV: 849 OBS MISSING DV: 371 OBS
VARIABLES COEF. STD.
ERR.
P>T COEF. STD.
ERR.
P>Z COEF. STD.
ERR.
P>Z
MOTIVATION
INC_RESOURCE -0.0229 0.0076 0.003 -0.0206 0.0097 0.036 -0.0252 0.0066 0.0000
MARKET GROWTH 0.4155 0.0912 0 0.3265 0.1069 0.003 0.4369 0.0922 0.0000
MARKET DEMAND 0.0001 0 0.012 0.0001 0 0.043 0.0001 0.0000 0.0020
PRE-THREAT PRICE 0.0465 0.0128 0 0.0626 0.0143 0 0.0414 0.0113 0.0000
PRE-THREAT DELAY 0.0061 0.0082 0.454 0.0074 0.0089 0.406 0.0071 0.0075 0.3420
INC_ROUTE_IMPORTANCE -1.4946 0.1641 0 -1.4411 0.1736 0 -1.4969 0.1631 0.0000
CAPABILITY
PE_RESOURCE 0.098 0.0099 0 0.0822 0.012 0 0.1008 0.0098 0.0000
PE_SIZE 0.0095 0.0024 0 0.0116 0.0026 0 0.0088 0.0024 0.0000
FUELPRICE -0.17 0.0276 0 -0.1589 0.03 0 -0.1776 0.0256 0.0000
CONTROLS
MARKET DISTANCE 0.0173 0.0042 0 0.0162 0.0047 0.001 0.0166 0.0042 0.0000
DISTANCESQ -0.0007 0.0002 0 -0.0007 0.0002 0 -0.0007 0.0002 0.0000
MMC 0.8508 0.1065 0 0.817 0.1255 0 0.8387 0.1126 0.0000
INCUMBENT_NUM 0.274 0.0369 0 0.3226 0.0353 0 0.2557 0.0357 0.0000
LEISURE 0.1249 0.016 0 0.1342 0.017 0 0.1142 0.0146 0.0000
HUB 0.001 0.0178 0.955 0.0028 0.0207 0.894 0.0015 0.0180 0.9340
LOAD_FACTOR -0.1493 0.0328 0 -0.1479 0.0336 0 -0.1281 0.0305 0.0000
INTERACTION INC_RESOURCE*MMC 0.2153 0.0624 0.001 0.1994 0.0692 0.004 0.2159 0.0581 0.0000
100
TABLE A.4: Robustness Check-Distribution Assumption
*p < .10, **p < .05, ***p < .01
ALL MISSING = REAL ALL MISSING = BLUFF
Model 5 (4245 Obs) Model 6 (4245 Obs)
Variables Coef. Std. Err. P>z Coef. Std. Err. P>z
MOTIVATION
Inc Resource -0.030*** 0.008 0.000 -0.026*** 0.006 0.000
Market Growth 0.345*** 0.094 0.000 0.447*** 0.074 0.000
Market Demand 0.000 0.000 0.102 0.000*** 0.000 0.000
Pre-threat Price 0.095*** 0.011 0.000 0.015 0.010 0.119
Pre-threat Delay 0.007 0.008 0.407 0.012 0.007 0.070
INC Route Importance -1.373*** 0.165 0.000 -1.331*** 0.145 0.000
CAPABILITY
PE Resource 0.057*** 0.009 0.000 0.086*** 0.009 0.000
PE Size -0.001 0.004 0.859 -0.001 0.003 0.672
FuelPrice -0.169*** 0.029 0.000 -0.172*** 0.023 0.000
CONTROLS
Market Distance 0.013*** 0.004 0.002 0.021*** 0.003 0.000
DistanceSQ -0.001*** 0.000 0.000 -0.001*** 0.000 0.000
MMC 0.454*** 0.108 0.000 0.674*** 0.100 0.000
Incumbent num 0.378*** 0.032 0.000 0.196*** 0.032 0.000
Leisure 0.114*** 0.016 0.000 0.103*** 0.013 0.000
Hub 0.009 0.018 0.599 -0.014 0.016 0.373
Load factor -0.183*** 0.031 0.000 -0.108*** 0.027 0.000
INTERACTION Inc Resource*MMC 0.232*** 0.064 0.000 0.240*** 0.056 0.000
101
TABLE A1.5: Robustness Check-Distribution Assumption (Cont’d)
70% ARE BLUFF 50% ARE BLUFF 30% ARE BLUFF
Variables Mean STD Min Max Mean STD Min Max Mean STD Min Max
MOTIVATION
Inc_resource -0.027
0.003 -0.037
-0.019
-0.028
0.003 -0.038
-0.019
-0.027
0.003 -0.034
-0.019
Market Growth 0.417 0.039 0.295 0.536 0.395 0.042 0.267 0.525 0.417 0.038 0.308 0.526
Market Demand 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Pre-threat Price 0.039 0.005 0.022 0.055 0.055 0.006 0.030 0.073 0.039 0.005 0.024 0.056
Pre-threat Delay 0.010 0.003 0.000 0.020 0.009 0.003 -
0.001
0.020 0.010 0.003 0.001 0.022
INC_Route_Importance -
1.342
0.063 -
1.529
-
1.089
-
1.349
0.071 -
1.549
-
1.125
-
1.342
0.064 -
1.558
-
1.127
CAPABILITY
PE_Resource 0.078 0.002 0.071 0.086 0.072 0.003 0.063 0.081 0.078 0.003 0.070 0.086
PE_Size -
0.001
0.001 -
0.005
0.003 -
0.001
0.001 -
0.006
0.003 -
0.001
0.001 -
0.005
0.004
FuelPrice -
0.170
0.013 -
0.209
-
0.128
-
0.170
0.015 -
0.219
-
0.123
-
0.172
0.014 -
0.209
-
0.127
CONTROLS
Market Distance 0.019 0.002 0.013 0.024 0.017 0.002 0.012 0.025 0.019 0.002 0.012 0.025
DistanceSQ -
0.001
0.000 -
0.001
-
0.001
-
0.001
0.000 -
0.001
0.000 -
0.001
0.000 -
0.001
0.000
MMC 0.606 0.040 0.488 0.731 0.562 0.042 0.399 0.696 0.607 0.040 0.485 0.721
Incumbent_num 0.250 0.016 0.193 0.299 0.287 0.017 0.235 0.344 0.251 0.016 0.205 0.310
Leisure 0.106 0.006 0.090 0.124 0.108 0.007 0.091 0.127 0.106 0.006 0.086 0.127
Hub -
0.007
0.007 -
0.030
0.013 -
0.002
0.008 -
0.029
0.026 -
0.007
0.007 -
0.029
0.016
Load_factor -
0.130
0.010 -
0.159
-
0.090
-
0.146
0.011 -
0.178
-
0.111
-
0.130
0.010 -
0.163
-
0.101
INTERACTION Inc_Resource*MMC 0.237 0.020 0.167 0.299 0.236 0.023 0.172 0.304 0.237 0.021 0.174 0.302
102
APPENDIX B: FULL LIKELIHOOD FUNCTION IN THE SPLIT-
POPULATION HAZARD MODEL
The split-population hazard model uses a mixture distribution: a logistic regression
estimates the proportion of new entrants that ‘never’ exit and a hazard regression estimate
the exit timing of new entrants that do exit a market at some point throughout the
observation period. This model enables us to investigate simultaneously the effect of
marketing covariates on the exit likelihood irrespective of time (incidence or ‘logit part’ of
the model), and the effect of marketing covariates on the time-to-exit for those challengers
that do exit the market (latency or ‘hazard part’ of the model).
Let t be a random variable denoting time-to-exit or survival time, with a cumulative
probability distribution F(t), hazard rate h(t), and survival function S(t) = 1 - F(t). Let Y =
1 denote an incidence, and Y = 0 no incidence, of the event of interest, δ = 1 indicates an
exit was observed in the data (non-censored observation) and δ = 0 indicate no exit was
observed (censored observation). Hence, there are three types of challengers (note that δ =
1 and Y = 0 simultaneously is impossible): (i) those that may and do leave the market
during our observation period (δ = 1, Y = 1); (ii) those that are likely to leave the market,
but outside our observation period (δ = 0, Y = 1); and (iii) those that are unlikely to leave
the market, even in the future (δ = 0, Y = 0). Essentially, split-population models use the
functional form of the hazard function to help distinguish
103
between the last two types of observations: using data on the probability of exit and
time-to-exit for low-cost carriers that do exit the market at some point throughout the
observation period, the model imputes the probability of exit and time-to-exit for carriers
for which no exit is observed. The survival function maps the probability that the survival
time is greater than or equal to t, and is given by:
S(t | X(t), Z) = [π(Z)S(t | Y = 1, Z, X(t))] + [1 - π(Z)] (1)
where Z and X(t) denote the vector of covariates that affect exit likelihood irrespective of
time and the vector of covariates that affect time-to-exit, respectively. Y = 1 denotes an
incidence of the event of interest (i.e., an exit), π(Z) is the probability of exit irrespective
of time, and S(t|Y=1, Z, X(t)) is the survival function (conditional on exit). If all firms exit
the market (all observations are ‘non-cured’), the model reduces to the standard survival
model, i.e. π = 1 (and 1-π = 0). Notice that Y = 1 occurs with probability π, and thus Y = 0
(a challenger that will not exit the market, i.e., a ‘long-term survivor’ in medical jargon)
occurs with probability 1- π. The likelihood for observation i (a market or route with a new
low-cost entrant or challenger) in quarter j is thus:
Li,j(b,β,βT) = [π(zi)h(tj | Y = 1, zi, xi(tj))S(tj | Y = 1, zi, xi(tj))]yiδi,j × [1 - π(zi)]
(1 - yi)(1 -
δ(i,j)) × [π(zi)S(tj | Y = 1, zi, xi(tj))]yi(1 - δ(i,j)) (2)
where δi,j denotes the quarter-specific censoring indicator for observation i.32 By
rearranging terms (note that the survival function S(.) is common to the first and last
components) and applying logs, the full log-likelihood function is given by (to facilitate
32 When Y = 1 and 𝛿 = 1, the exponents of the last two components become zero and the likelihood is
reduced to the first component only; when Y = 1 and 𝛿 = 0, the exponents of the first two components
become zero and the likelihood is reduced to the last component only; when Y = 0 and 𝛿 = 0 the exponents
of the first and last components become zero and the likelihood is reduced to the second component only.
104
the reading we omit the conditional Y=1) the sum of the incidence and latency log-
likelihoods, i.e.
llinc(b|z) = log (∏ [1 − π(zi)]1−yini=1 π(zi)
yi) = ∑ (1 − yi)ni=1 log[1 − π(zi)] +
yilog (π(zi)),
and
lllat(b, β, βT|z, x(t)) = log(∏ ∏ h[(tj|zi,xi(tj))]yiδi,jS[(tj|zi,xi(tj))]
yim
j=1
n
i=1)
= ∑ ∑ yiδi,jlog h[(tj|zi,xi(tj))] + yilog [S(tj|zi,xi(tj))]mj=1
ni=1 , (3)
respectively.33
Split Population Hazard Model (Logit part specification)
We specify the logit part of the model as a function of pre-entry average market conditions
because they reflect the type and level of required resources that determine market survival
in general, i.e., irrespective of time (see Helfat and Lieberman 2002). Specifically, π(zi) on
route i is a function of incumbents’ pre-entry prices (IncPrePricei), service convenience,
measured by flight frequency during both non-peak (IncPreFreqi) and peak-time
(IncPrePeakFreqi), and service quality, measured by both on-time performance
(IncPreOTPi) and plane size (IncPrePlaneSizei), and is specified as follows (see e.g., Wei
and Hansen 2005):
log(π(zi)/ 1 - π(zi)) = γ0 + γ1IncPrePricei + γ2IncPreFreqi + γ3IncPrePeakFreqi +
γ4IncPreOTPi + γ5IncPrePlaneSizei + γ6ChllgPriceij
(4)
33 It is impossible to know, from observed data, whether a low-cost carrier will never exit a given route or is
just right-censored. In the unlikely case that all carriers would exit, the split-population model would
incorrectly identify some of them as being cured, i.e., never exit (see Jaggia 2011). This is more likely in
short datasets. Because our dataset leaves plenty of time for those carriers that entered routes long time ago
to exit them, We believe that a split-population model is more realistic than a hazard model that assumes
the data are right-censored.
105
where IncPrePricei, IncPreFreqi, IncPrePeakFreqi, IncPreOTPi, and IncPrePlaneSizei are,
respectively, incumbents’ pre-entry prices, service convenience measured by both non-
peak and peak-time flight frequency, and service quality measured by both on-time
performance (OTP) and plane size (Wei and Hansen 2005). All measurements are averages
over eight pre-entry quarters. We also control for the challenger’s price ChllgPriceij. Note
that omitted variables in the probability of exit (logit part) are assumed to be independent
of omitted variables in the time-to-exit. While this may not be a particularly realistic
assumption (unobserved characteristics that make challengers less likely to exit are
probably the ones that make challengers less likely to exit sooner), it is less problematic
than the stronger assumptions of both models (see Goldhaber, Krieg, and Theobald 2014
for a similar argument). A hazard model would assume there is no error in the probability
of exit (every challenger is assumed to exit) and a logit model would assume there is no
error in the time-to-exit (as the outcome is binary).
106
APPENDIX C: 2SRI METHOD
A two-stage residual inclusion estimation method (2SRI) is an extension of the popular
two-stage least squares (2SLS). The 2SLS is not consistent for nonlinear models, whereas
the 2SRI estimator is (Terza, Basu, and Rathouz 2008). The first stage equation regresses
the incumbents’ prices on a set of exogenous variables and an instrumental variable (IV),
which must be correlated with prices but not with the new entrant’s time-to-exit. In the
second stage of 2SRI prices are not replaced: the first-stage residuals are instead included
as an additional variable. Following the footsteps of previous studies of price elasticity that
controlled for price endogeneity in the airline industry (Lurkin et al. 2017; Mumbower,
Garrow, and Higgins 2014), we use the number of connecting passengers (ConnPass) as
an IV.
In the airline industry, one-stop routes, say A-B-C or B-C-D, and non-stop routes, say B-
C, are two distinct types of markets facing a different demand: while one-stop routes serve
connecting passengers (those flying from A to B and then from B to C or from B to C and
then from C to D), non-stop routes do not. Typically, major incumbent carriers serve one-
stop routes and therefore carry not only direct passengers but also connecting passengers.
Low-cost carriers, in turn, serve non-stop routes with virtually no connecting
107
passengers.34 A large number of connecting passengers, i.e., a high demand in A-B-C or
B-C-D routes, will significantly increase the demand for incumbents’ B-C route and thus
reduce the incumbents’ cost for each available seat mile in that route and thereby its fares
(Shaw 2007)35. On the contrary, and by definition, a large number of connecting passengers
should not affect a challenger’s non-stop B-C route. In other words, the number of
connecting passengers, ConnPass, is likely to be correlated with incumbents’ prices (in
one-stop routes) but not with the challenger’s market-share and profitability (in non-stop
routes), nor the challenger’s exit likelihood or time-to-exit.
We find that ConnPass satisfies the exclusion restriction and relevance criteria
(Cameron and Trivedi 2005), i.e. (a) it is not directly correlated with the challenger’s
hazard rate and (b) it is sufficiently correlated with our price variable. Dixit and
Chintagunta (2007) also suggest that a good instrument for the incumbents’ prices in the
context of airlines market exit is time-varying and airline-market specific, and ConnPass
meets these two criteria. Moreover, we test if the ConnPass instrument is strong enough
by regressing incumbents’ prices on ConnPass. Following Ebbes et al. (2016) we estimate
two first-stage models. The first one includes ConnPass and all other independent variables
in the main regression equation, and the second one includes only exogeneous variables
and excludes ConnPass. We find that the incremental F-statistic between these two models
is significantly greater than 10 (ΔF = 25), which indicates that the instrument is a strong
one (Ebbes et al. 2016).
34 The only exception is AirTran airlines that used a hub-and-spoke business model, and so we dropped
AirTran even at the expense of losing roughly 20% of the observations. We also estimated the IV model
using all observations and the conclusions regarding the endogeneity concerns remain valid. 35 notice that more than 65% of an airline’s costs, e.g., fuel costs, crew salaries, airport landing fees, aircraft
leasing fees, are independent of the number of passengers on a plane.
108
Having a proper instrument, we first test the null hypothesis that price can be treated as an
exogenous regressor using the t-statistic associated with the residual included in the second
stage. Following Lurkin et al. (2017), the insignificant t-statistic indicates that price should
be considered as an exogenous variable and endogeneity is not a real concern. In our case,
the residual coefficient is not significant (t = .74, p > .10), suggesting that price is not
endogenous
109
APPENDIX D: VARIABLE MEASUREMENTS
MMC Calculation:
as follows. For challenger c, in route i, we count all common routes with incumbents over
all routes in quarter j and then divide the challenger’s total contact by (n - 1), where n is
the number of incumbents (including the challenger) in route i. Finally, we standardized
the average count by the number of markets served by the challenger in quarter j
Route Importance:
We employ the measure developed by Dunn (2008): for each route, the network importance
measure is determined by the number of non-stop markets that originate from the two
endpoints (excluding the non-stop route to the city being considered) divided by its network
size. For instance, if, in a route between city “O” and city “D”, an LCC has five non-stop
routes out of “O” and six non-stop routes out of “D”, and it serves 100 routes within its
network, then the network centrality (importance) of route O-D is [(5+6)-2] / 100 = .09.
110
APPENDIX E: TYPES OF MISSINGNESS
Missing at random vs. missing not at random
There are three types of missing data depending on the mechanisms that may generate
them: (1) missing at random (MAR) data refers to the case when the “missingness” in the
dependent variable (Yi ) does not depend on its value, but may depend on the value of other
variables. In other words, after controlling for other variables, the probability of the missing
Yi is not related to the value of Yi; (2) missing completely at random (MCAR) data are a
special case of (MAR), which means that the probability of missing data on Yi does not
depend on Yi value or the values of any other variable in the data set; (3) if the MAR
assumption is violated then the missing data are not at random (MNAR) and the
missingness mechanism and data are nonignorable. In this latter case, the reason for
missingness often depends on the missing values themselves. For instance, nonresponse in
an income survey may be related to an unobserved income. A missing data is ignorable if
(a) the missingness mechanism is at random (MAR), and (b) the parameters for a missing-
data generation are unrelated to the parameters to be estimated. MAR and “ignorability”
are often equivalent since assumption b is almost always satisfied.
111
MICE Steps
The chained equations algorithm cycles through incomplete variables one at a time,
drawing imputations from a series of univariate conditional distributions. At each
imputation round, missing values for a particular variable are drawn from a distribution
that conditions on all other variables, including filled-in variables from a previous step.
MICE follows the following steps to impute missing values (Azur, et al. 2011):
Step 1: Initially, a simple imputation is performed for every missing value for each variable
in the dataset. At this step, usually, missing values will be replaced by the mean of the
observed values. These imputed values can be thought of as a “place holder”
Step 2: The “place holder” – imputed values -- for one variable “X” are set back to missing.
Step 3: The observed values from the variable “X” in Step 2 are regressed on the other
variables in the imputation model, which may or may not consist of all of the variables in
the dataset. In other words, “X” becomes the dependent variable in a regression model and
all the other variables are independent variables in the regression model. These regression
models operate under the same assumptions that one would make when performing linear,
logistic, or Poison regression models outside of the context of imputing missing data.
Step 4: The missing values for “X” are then replaced with predictions (imputations) from
the regression model. When “X” is subsequently used as an independent variable in the
regression models for other variables, both the observed and these imputed values will be
used.
Step 5: For variables that have missing observations, repeat steps 2–4 for a number of
rounds to create several complete datasets. At the end of each round, all of the missing
112
values will be replaced with predictions from regressions that reflect the relationships
observed in the data.